Research ArticleStructural Analysis and Total Coal Demand Forecast in China

Qing Zhu12 Zhongyu Zhang12 Rongyao Li3 Kin Keung Lai24

Shouyang Wang5 and Jian Chai25

1 School of Finance and Economics Xirsquoan Jiaotong University Xirsquoan 710061 China2 International Business School Shaanxi Normal University Xirsquoan 710119 China3 College of Economics and Management Southwest University Chongqing 400715 China4College of Business City University of Hong Kong Tat Chee Avenue Kowloon Hong Kong5 National Center for Mathematics and Interdisciplinary Science Chinese Academy of Sciences Beijing 100190 China

Correspondence should be addressed to Jian Chai chaijian0376126com

Received 7 March 2014 Accepted 16 May 2014 Published 5 June 2014

Academic Editor Chuangxia Huang

Copyright copy 2014 Qing Zhu et alThis is an open access article distributed under theCreativeCommonsAttribution License whichpermits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

Considering the speedy growth of industrialization and urbanization in China and the continued rise of coal consumption thispaper identifies factors that have impacted coal consumption in 1985ndash2011 After extracting the core factors the Bayesian vectorautoregressive forecast model is constructed with variables that include coal consumption the gross value of industrial outputand the downstream industry output (cement crude steel and thermal power) The impulse response function and variancedecomposition are applied to portray the dynamic correlations between coal consumption and economic variables Then foranalyzing structural changes of coal consumption the exponential smoothing model is also established based on division ofseven sectors The results show that the structure of coal consumption underwent significant changes during the past 30 yearsConsumption of both household sector and transport storage and post sectors continues to decline consumption of wholesale andretail trade and hotels and catering services sectors presents a fluctuating and improving trend and consumption of industry sectoris still highThegross value of industrial output and the downstream industry output have beenpromoting coal consumption growthfor a long time In 2015 and 2020 total coal demand is expected to reach 274627 and 404168million tons of standard coal in China

1 Introduction

As the basic source of energy coal has facilitated the rapiddevelopment of the Chinese economy and has had a positiveeffect on the stability of the market economy Therefore inorder to avoid indiscriminate production effectively improv-ing the efficiency of the coal industry and ensuring nationalenergy security and accurate prediction of coal demand arenecessary

In the past scholars have established a variety of energydemand forecasting models for different countries andregions as well as different kinds of energy Traditional meth-ods such as time series regression econometric decomposi-tion unit root test and cointegration ARIMA (autoregressiveintegrated moving average) and input-output as well assoft computing techniques such as fuzzy logic GA (geneticalgorithm) and ANN (artificial neural networks) are being

extensively used for demand side management Support vec-tor regression ACO (ant colony) and PSO (particle swarmoptimization) are new techniques being adopted for energydemand forecasting Bottom up models such as MARKAL(acronym for MARKet ALlocation) and LEAP (long-rangeenergy alternatives planningmodel) are also being used at thenational and regional level for energy demand management[1]

For foreign energy consumption Ediger and Akar[2] considered that estimated economic and demographicparameters usually deviate from the realizations Time-seriesforecasting appears to give better results so they usedthe ARIMA and SARIMA (seasonal ARIMA) methods toestimate the future primary energy demand of Turkey from2005 to 2020 Mohr and Evans [3] considered the modelof worldwide coal production developed for three scenar-ios (Hubbert Linearisation method scenario reserves plus

Hindawi Publishing CorporationDiscrete Dynamics in Nature and SocietyVolume 2014 Article ID 612064 10 pageshttpdxdoiorg1011552014612064

2 Discrete Dynamics in Nature and Society

cumulative production scenario best guess scenario) Theultimately recoverable resources (URR) estimates used in thescenarios ranged from700Gt to 1243GtUnler [4] proposed amodel to forecast the energy demand of Turkey until 2025 byusing PSO-based energy demand forecasting (PSOEDF) andmade a comparison with the ant colony optimization (ACO)energy demand estimation model The result showed thatthe former algorithm had better accuracy Lee and Chiu [5]applied a newly developed panel smooth transition regressionmodel with the error-correction term (PSECM) to estimatethe nonlinear relationships among energy consumption realincome and real energy prices for 24 OECD countries

For domestic energy consumption various approachesto forecasting have been used in the literature includingcointegration and error correction models [6] and demandequations [7] In 2002 Yu and Zhu [8] proposed animproved hybrid algorithm called PSO-GA (particle swarmoptimization-genetic algorithm) for energy demand forecast-ing in China with higher precision compared with singleoptimization methods such as GA PSO or ant colony opti-mization andmultiple linear regressions Results of this studyshow that Chinarsquos energy demandwill be 470 billion tons coalequivalent in 2015 In the same year they also proposed mix-encoding particle swarm optimization and radial basis func-tion (MPSO-RBF) network-based model to forecast Chinarsquosenergy consumption until 2020 based on GDP populationproportion of industry in GDP urbanization rate and shareof coal energy for the period from 1980 to 2009 [9] Zhanget al [10] forecast transport energy demand for 2010 2015and 2020 based on partial least square regression (PLSR)method under two scenarios by analyzing gross domesticproduct (GDP) urbanization rate passenger-turnover andfreight-turnover for the period of 1990ndash2006 Crompton andWu [11] applied the Bayesian vector autoregressive method-ology to forecast Chinarsquos energy consumption and to discusspotential implicationsThe results of this paper suggested thattotal energy consumption should increase to 2173 MtCE in2010 an annual growth rate of 38 which is slightly slowerthan the average rate in the past decade due to structuralchanges in the Chinese economy

Among them some of the latest techniques such asBayesian vector autoregression (BVAR) support vector re-gression ant colony and particle swarm optimizationmodelsare being used in energy demand analysis Chai et al [12]established a VARX (vector autoregressivemodel with exoge-nous variables) model of crude oil market structure to studythe effect of every variable on oil price by screening variableswith price supply demand of oil dollar index and Chinaoil net import as the endogenous variables and reservespeculative factors as the exogenous variablesThen based onthis VARX model and the Bayes theory a MSBVAR (Markovswitching bayes vector autoregression) model is establishedto identify and analyze the structural changing of oil pricesystem structure within the study period BVAR model andGranger-causality are applied to study growth in energydemand and the relationship between energy consumptionand real gross domestic product per capita in selected fewCaribbean countries [13] Bayesian neural network approachis used for short term electric load forecasting [14 15]

The BVAR model can avoid the rigid inclusionexclusionrestrictions of VAR models by allowing inclusion of manycoefficients while simultaneously controlling the extent ofmathematical expectation and standard deviation to whichthey can be influenced by the data This reduces the spuriouscorrelations captured by themodel thereby reducing forecasterror and improving forecasting performance

On the basis of the background and demand mentionedabove this paper attempts to analyze structural changes ofcoal consumption during the past 30 years and forecast futurecoal demand in China up to 2020 This paper includes sevenparts introduction describing the realistic background andthe academic background for research topic variables selec-tion and data processing screening the core effect factors fortotal coal demand methodology introducing VAR BVARand ETS models forecast of total coal consumption basedon VAR models applying the impulse response functionand variance decomposition to portray the dynamic corre-lations between coal consumption and economic variablesand forecasting total coal consumption by BVAR modelsstructural analysis of total coal consumption establishingETS models based on coal consumption of seven sectors toanalyze structural changes of coal consumption and forecasttotal coal consumption results and discussion comparingforecast results of total coal consumption by BVAR modelsand ETS models conclusions

2 Variables Selection and Data Processing

Many factors that affect future demand for coal such asdomestic and international price of coal the national policyof saving energy and reducing consumption efficiency ofresource utilization coal consumption of downstream indus-tries railway and highway capacity for coal transportationtechnology development urbanization industrial growthindustrial structure the supply and price of other alternativeenergy sources (such as oil natural gas hydropower nuclearpower etc) coal production cost (pollution control costscoal mine safety production costs exit cost etc) and con-sumer habits

Based upon the China Statistical Yearbook China EnergyStatistical Yearbook China Coal Industry Yearbook andWIND Database this paper has identified the followinginfluence factors of coal demand from 1985 to 2011 (1)unit of coal consumption is ten thousand tons of standardcoal (2) since most coal is consumed in industry andcontribution of the secondary industry to gross domesticproduct (GDP) fluctuates sharply gross industrial output (100million yuan) and industrial structure () are consideredimportant factors that affect coal consumption (3) electricpower steel building materials and chemical industry arefour major downstream industries that consume coal sooutput of thermal power (100 million kilowatt hour) crudesteel (10000 tons) and cement (10000 tons) are also factorsused to forecast coal consumption (4) effect of internationalprices and supply position of different substitute sources ofenergy on domestic coal usage and consumption such ascoal price (Newcastle in AustraliaKen blah FOB) oil price

Discrete Dynamics in Nature and Society 3

(Europe Brent Spot Price FOB) natural gas price (Louisianaspot prices) and producer price indices for domestic indus-trial products by sector (power coal and petroleum) (5)per capita GDP (yuan) and urbanization rate () data arecollected since the improvement of consumer capacity andchange of urban structure affect the energy consumption percapita in China

First of all in order to eliminate the heteroscedasticity ofthe economic time series data and to linearize its trend theabove variables are put into the natural logarithms Then toprevent false regression leading to an invalid conclusionADF(Augment Dikey-Fuller) unit root test and cointegration testare used to examine stationarity and long-run equilibriumrelationship of logarithmic time series respectively Finallyonly the following 5 groupsrsquo data meet the conditions ofthe VAR model coal consumption (COAL) gross value ofindustrial output (IND) output of cement industrial products(CEM) output of crude steel industrial products (STE) andoutput of thermal power industrial products (POW)

3 Methodology

31 VAR A vector autoregressive (VAR)model based on sta-tistical properties of the data is established without exertinga theory of a priori constraints for the data mechanism so itis an unstructured time series model not a structural econo-metric model with a theoretical foundation Specifically alleconomic variables are considered as endogenous variablesin VARmodel through multistage lag regression to estimatetheir relationships and for establishing forecast models

In 1980 Sims [16] introduced the VAR model and pro-moted its application for dynamic analysis of the economicsystem extensively VAR model is used to forecast inter-connected time series system and analyze dynamic shocksfrom stochastic disturbances to variables so as to explain theinfluence of various economic shocks on economic variables

An unrestricted vector autoregression (UVAR) modelcontaining 119899 time series variables and a lag length of 119901 hasthe following general form

119910119905= 119888 + 119860

1119910119905minus1

+ sdot sdot sdot + 119860119901119910119905minus119901

+ 120583119905

120583119905sim 119894119894119889119873 (0 120590

2) 119905 = 1 2 119899

(1)

where 119910119905= (1199101 1199102 119910

119899)1015840 is an (119899times1) vector of endogenous

variables 119888 is an (119899 times 1) vector of intercept terms 119860119894 119894 gt 1

are (119899 times 119899) coefficient matrices and 120583119905is an (119899 times 1) vector of

the independent normal random error vectorBased on the VAR model the dynamic relationship

between the variables can be analyzed by impulse responsefunction and variance decomposition

Impulse response function describes the effect on thecurrent and future variables from information shocks of arandom disturbance which equals a standard deviation andvividly depicts the path changes of the dynamic interactionbetween the variables and tests the intensity and durationof the impact of economic variables on coal consumptionThis function implies that decomposition does not have tocompletely depend on the order of the variables in the VAR

system so as to improve the stability and reliability of theestimation results

Variance decomposition method decomposes meansquare error (MSE) of the forecast system into contributionsof each variable facilitating examination of MSE decompo-sition of any endogenous variables and measurement of therelative importance of random disturbances for the variablesby comparing the variance contribution rate In this paperthrough the variance decomposition method the role of theeconomic variables fluctuation in coal consumption growthcan be determined

32 BVAR In (1) the vector 119888 contains 119899 intercept termsand each matrix 119860

119894contains 1198992 coefficients hence with the

increase of variables and lag 119899 + 1199011198992 coefficients must beestimated which increase exponentially with the numberof variables in the system A major problem in estimationof VAR models when 119901 is large is over-parameterizationwhere too many coefficients must be estimated relative tosample size This can lead to an overfitting problem the largenumber of coefficients in unrestricted VAR models tendsto fit the data unrealistically well while performing poorlyin out-of-sample forecasting due to the effects of spuriouscorrelations in the data set Over-fitting of the data candistort the long-run relationships between variables in themodel and inflate coefficient values on distant lags due to lowdegrees of freedom [17] In addition VAR model ignores thea priori information and presets the same importance for allestimated parameters which can lead to the wrong modelThere are many ways to overcome the over-fitting problemBy applying certain constraints on the parameters such asreducing the lag length or removing some of the variablesin an individual equation can solve the problem of lowerdegrees of freedom effectively But from the view of Bayesianthis means that forecasters believe that the probability ofremoving lags whose coefficient is zero is 100 But it isimpossible to know whether this constraint is established

An alternate approach is the BVAR method proposed byLitterman [18 19] Doan et al [20] Sims [21] and Sims andZha [22] The BVAR approach modifies the OLS estimatesof (1) by treating all coefficients as random variables aroundtheir Bayesian prior mean such that the model has theflexibility to impose these priors to varying degrees on thecoefficient estimates

33 ETS ETS technology appeared and was used in the1950s After decades of development it becamemature How-ever selection of ldquooptimumrdquo model from different modelsfor use began few years ago Hyndman and Khandakar[23] summed up exponential smoothing models and dividedthem into fifteen types For more details see Table 1

Error Trend and Seasonal of ETS model respectivelyrepresent error term trend term and season term in which(Trend Seasonal) group includes fifteen models listed inTable 1 Residual error term is classified as overlapping forms(addition forms) and multiplication forms On the assump-tion of 119910

119905= 120583119905+ 120576119905 it is an additive errors model On the

assumption of 119910119905= 120583119905(1 + 120576

119905) it is a multiplication errors

model Under the circumstances of consideration of different

4 Discrete Dynamics in Nature and Society

Table 1 Classification and summary of ETS models

Trend component Seasonal componentN (none) A (additive) M (multiplicative)

N (none) N N N A N MA (additive) A N A A A MAd (additive damped) Ad N Ad A Ad MM (multiplicative) M N M A M MMd (multiplicativedamped) Md N Md A Md M

error forms the above fifteen models can be extended tothirty types From the forecast results alone addition formor multiplication form hardly influence residual error termHowever for different sampled data various residual errorforms seem superior or inferior Because dividing by 0 isinvolved attention should be paid to select different additionor multiplication forms for residual error trend term orseason term When samples are all positive values usingresidual error multiplication form presents an advantageHowever when samples are zero value or negative valuemultiplication form model is not applicable

4 Forecast of Total Coal Consumption byBVAR Models

In order to analyze the dynamic relationship between coalconsumption and economic variables 5 d VARmodel is builtbased on (1) According to the lag length criteria (likelihoodratio final prediction error AIC SC and Hannan-Quinninformation criterion) VAR (2) is the most appropriatemodel VAR models are set up as follows

[[[[[

[

LNCOALLNINDLNCEMLNSTELNPOW

]]]]]

]

=

[[[[[

[

44768

minus200313

31506

minus24537

23387

]]]]]

]

+

[[[[[

[

06785 01126 minus01174 00534 04913

01305 05586 14616 minus04848 06649

minus06834 00470 10756 00198 06238

04434 minus02351 00606 08957 00600

minus02571 00139 01578 03482 09504

]]]]]

]

times

[[[[[

[

LNCOAL (minus1)LNIND (minus1)

LNCEM (minus1)

LNSTE (minus1)LNPOW (minus1)

]]]]]

]

00

05

10

15

00 05 10 15

minus05

minus10

minus15

minus05minus10minus15

Figure 1 Inverse roots of AR characteristic polynomial

+

[[[[[

[

minus03453 minus01248 minus00067 01699 minus02082

20973 minus04018 minus03793 minus05128 minus04853

03309 minus00134 minus07911 minus00454 03111

minus02974 00022 minus00554 minus02683 07020

minus00600 minus00030 minus01852 minus01937 00594

]]]]]

]

times

[[[[[

[

LNCOAL (minus2)LNIND (minus2)

LNCEM (minus2)

LNSTE (minus2)LNPOW (minus2)

]]]]]

]

+

[[[[[

[

1205761

1205762

1205763

1205764

1205765

]]]]]

]

(2)

The R-squared of the models are 09981 09951 09966and 09986 which indicate good simulation results At thesame time the reciprocals of all roots locate inside the unitcircle when calculating the AR characteristic polynomial(Figure 1) which shows that the established VAR (2) modelis stable That is to say when a variable changes in the model(ie to generate a shock) it leads to changes in other variablestoo But as time goes on the effect gradually disappears InFigure 1 the horizontal axis represents the real number axisthe vertical axis represents the imaginary numbers axis theunit circle is a circle of radius 1 and center at the origin thepoints in the unit circle represent roots of the characteristicequation on VAR (2) model

Based on the VAR model the dynamic relationshipbetween the variables can be analyzed by impulse responsefunction and variance decomposition

In Figure 2 the horizontal axis represents the lag periodsof shock from innovation (unit years) the vertical axisrepresents the response degree from the dependent variablesto the explaining variables the solid line is the calculatedvalue of the impulse response function The lag period is setto 15 years

Discrete Dynamics in Nature and Society 5

000

001

002

003

004

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

minus001

minus003

minus002

minus004

LNCOALLNINDLNCEM

LNSTELNPOW

Figure 2 Responses of LNCOAL to LNCOAL LNIND LNCEMLNSTE and LNPOW

According to Figure 2 the line graph shows that when theexplaining variables (coal consumption gross value of indus-trial output cement output crude steel industrial output andthermal power output) are shocked from innovation respec-tively how does the dependent variable (coal consumption)respond during the future 15 periods That is to say whenthese five factors change respectively we can forecast howmuch and how long will coal consumption respond

Firstly coal consumption volatility responds strongly toits own shock which is shown by the blue line and inthe first 15 periods it shows a fluctuating trend Early asa strong positive response in the second period it reachesthe peak at 00310 and then plunges to the lowest point atminus00145 in the eighth period From this point onwards theperiod between 9 and 15 experiences a gradual growth untilthe numbers become positive response again This showsthat coal consumption and its lagging values have a strongcorrelation Secondly after suffering a positive shock grossvalue of industrial output only brings positive response tocoal consumption in the second period and then it quicklydrops to a negative response to the lowest point of minus00379in the sixth period Afterwards it rises to zero slightly Thisindicates that the increase of the gross value of industrialoutput immediately leads to an increase in coal consumptionbut in the long term it reduces consumption The reasonsmay be that more output provides more capital for inter-nal restructuring and technological progress thus reducingcoal consumption costs and increasing profits Thirdly coalconsumption volatility responds modestly to cement outputshocks in the early stage From the fifth period to theeleventh period the response shows a large fluctuation fromnegative to positive Fourthly coal consumption volatilityhas increasing positive response to crude steel industrialoutput shocks gradually over time and in the seventh periodthat reaches the peak of 00219 Fifthly a positive shockon thermal power output brings positive response to coalconsumption which has weak intensity and stable tendencyThese results indicate that the development of downstream

0

20

40

60

80

100

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

LNCOALLNINDLNCEM

LNSTELNPOW

Figure 3 Factors variance decomposition on the effect of LNCOAL

industries drives an increase in demand for coal whose long-term relation is in the same direction

In Figure 3 the horizontal axis represents the lag periodsof impact from innovation (unit years) the vertical axisrepresents the contribution rate to coal consumption from theeconomic variables

As shown in Figure 3 in the future 15 periods coalconsumption has the greatest impact during the first fiveperiods with its own contribution rate above 40 Since thenrates in the rest of the period level off at about 22 Thecontribution rate of the gross value of industrial output to coalconsumption increases steeply from the third period and thenexceeds all other variables at the sixth period and remainsstable and high between 51 and 53Thermal power industryhad the highest contribution in the second period but afterthat its contribution rates decline to 3-4 Contribution ratesof crude steel industry have obvious rise from the thirdperiod and make the largest contribution in the downstreamindustries between 12ndash14 The contribution rate of cementindustry has an obvious rise in the sixth period after thatthe contribution rate remains in the 7-8 level Therefore inthe long run industrial output has the biggest contributionto the changes of coal consumption in a short time coalconsumption have the biggest contribution to own changes

The system of priors commonly used in the specificationof BVAR models includes conjugate prior distribution max-imum entropy prior distribution ML-II prior distributionand multilayer prior distribution Different BVAR priordistributions have different impacts on the prediction resultsThis paper uses the 119877 software to forecast coal consumptionbased on three kinds of prior distribution in the MSBVARpackage Command is run as follows Rgt fcastlt -szbvar (coal119901 = 2 lambda 0 = 06 lambda 1 = 01 lambda 3 = 2 lambda4 = 025 lambda 5 = 0 mu 5 = 0 mu 6 = 0 prior = 0)Thereare three different prior distributions 0 = Normal-Wishartprior 1 = Normal-flat prior and 2 = flat-flat prior (ie akintoMLE)The forecasting results from 2012 to 2020 are shownin Table 4

6 Discrete Dynamics in Nature and Society

6000

8000

10000

12000

14000

16000

18000

050000

100000150000200000250000300000350000400000450000500000

1980

1985

1990

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

2018

2019

2020

Industry coal consumptionHousehold coal consumption

Coa

l con

sum

ptio

n (1

0000

tons

)

Coa

l con

sum

ptio

n (1

0000

tons

)

Figure 4 Changes of industry and household coal consumption

5 Structural Analysis of TotalCoal Consumption

A univariate ETS (exponential smoothing) forecast model isused to predict coal consumption of each sector in 2020 inthe forecast package for R Error Trend and Seasonal of ETSmodel denote error term (A M Z) trend term (N A M Z)and season term (NAM Z) respectively inwhich the groupincludes thirty models

From the coal balance sheet in China Statistical YearbookandWINDDatabase coal consumption data of seven sectorsare selected for the period 1980ndash2010 The balance equa-tion for coal consumption Total coal consumption (TO) =Agriculture Forestry Animal Husbandry Fishery andWaterConservancy (AFAFW) + Industry (IND) + Construction(CON) + Transport Storage and Post (TSP) +Wholesale andRetail Trades Hotels and Catering Services (WRHC) +OtherSectors (OS) + Household Consumption (HC) Unit is 10000tons

The finalmodels (Table 2) are selected from thirtymodelsby comparing information criterion (AIC BIC AICc) andforecast precision (mean error root mean square error meanabsolute error maximum permissible errors mean absolutepercentage error mean absolute error square) (Table 3)Command is run as follows R gt fcast lt -forecast(est(x))

As can be seen from the Table 4 the total coal consump-tion is the sum of coal consumption in seven sectors whichis 39728663 billion tons in 2015 This value is similar tothe goal in the development of coal industry in the twelfthfive-year plan which is 39 billion tons in 2015 By 2020Chinarsquos total coal consumption will reach 483 billion tonswith an annual growth rate of 436 of which industry andhousehold consumption account for the largest proportionIn Figure 4 the industry coal consumption was 438 milliontons 811 million tons 1278 billion tons and 296 billiontons in 1980 1990 2000 and 2010 respectively and it ispredicted to reach 4669 billion tons in 2020 accountingfor 9667 of total coal consumption Therefore to reducecoal consumption in China industrial energy conservationand emissions reduction are the fastest and most effective

breakthrough points including the adjustment of industrialstructure and optimization of technology

However as the second largest main channel the propor-tion of household consumption is decreasing year by yearFigure 4 shows that household consumption was up to 167million tons in 1990 which accounts for 1583 of the totalcoal consumption By 2020 household consumption willbe 092 million tons accounting for 189 of the total coalconsumption In the field of household energy consumptionas the diversity of the type of energy and development oftechnology electricity and heating oil are more available andconvenient than coal for residents so less and less coal is usedin the household rapidly This is beneficial to environmentalprotection and nonrenewable resources saving

As shown in Figure 5 transport storage and post sectorswere the largest coal consumption sectors in 1980 but theybecame the lowest coal consumption sectors after thirty yearsin 2010 and their coal consumption will decline constantlyduring the next decade by predictionsThemain reason is thatcoal-fired steam locomotive engines have been replaced byinternal combustion and electric locomotives by the railwaysSo most of the raw coal has been replaced by fuel andelectricity at the same time Nowadays road transporta-tion mainly consumes gasoline and diesel oil waterwaystransportation mainly consumes fuel oil and diesel oil airtransport mainly consumes aviation kerosene Thereforealthough the development of transport storage and postsectors still consumes large amounts of energy the usage ofcoal is insignificant

As the pillar industry of national economy in our countryconstruction sector is also an energy-intensive industryCoal consumption increased from 556 to 71892 milliontons between 1980 and 2010 representing a relatively smallgrowth By forecasting the amount of coal consumption isstill increasing in the future (Figure 5) At the same time theconstruction sector does not have a high proportion of coalconsumption and accounted for only 023 in 2010 That isbecause there are various types of energies consumed by theconstruction industry including raw coal gasoline diesel oilfuel oil heat electricity and other petroleum products With

Discrete Dynamics in Nature and Society 7

300

600

900

1200

1500

1800

2100

2400

2700

1980

1985

1990

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

2018

2019

2020

AFAFWCONTSP

WRHCOS

Coa

l con

sum

ptio

n (1

0000

tons

)

Figure 5 Changes of other sectorsrsquo coal consumption

Table 2 Formulae for recursive calculations point forecasts and parametric estimated value on sector by ETS method

Variables Forms Models and parameters

AFAFW (M N N)119897119905= 120572119910119905+ (1 minus 120572) 119897

119905minus1 119910119905+ℎ|119905

= 119897119905

119910119905+1

= 119910119905+1|119905

(1 + 120576119905) 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 00001 119897 = 17937859 120590 = 01809

IND (A A N)

119897119905= 120572119910119905+ (1 minus 120572) (119897

119905minus1+ 119887119905minus1) 119887119905= 120573 (119897

119905minus 119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905+ ℎ119887119905

119910119905+1

= 119910119905+1|119905

+ 120576119905 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 09999 120573 = 02697119897 = 547821109 119887 = 93385498 120590 = 129103600

CON (A A N)119897119905= 120572119910119905+ (1 minus 120572) (119897

119905minus1+ 119887119905minus1) 119887119905= 120573 (119897

119905minus 119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905+ ℎ119887119905

119910119905+1

= 119910119905+1|119905

+ 120576119905 120576119905sim 119873119868119863 (0 120590

2) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 00009 120573 = 00009 119897 = 4483868 119887 = 87382 120590 = 536530

TSP (M M N)119897119905= 119886119910119905+ (1 minus 119886) 119897

119905minus1119887119905minus1

119887119905= 120573 (119897

119905119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905119887119905

ℎ

119910119905+1

= 119910119905+1|119905

(1 + 120576119905) 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 00001 120573 = 00001 119897 = 23050531 119887 = 09371 120590 = 01049

WRHC (M M N)119897119905= 119886119910119905+ (1 minus 119886) 119897

119905minus1119887119905minus1

119887119905= 120573 (119897

119905119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905119887119905

ℎ

119910119905+1

= 119910119905+1|119905

(1 + 120576119905) 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 00001 120573 = 00001 119897 = 7651436 119887 = 10328 120590 = 02562

OS (A N N)119897119905= 119886119910119905+ (1 minus 119886) 119897

119905minus1 119910119905+ℎ|119905

= 119897119905

119910119905+1

= 119910119905+1|119905

+ 120576119905 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 08917 119897 = 11487978 120590 = 4978256

HC (M N N)119897119905= 119886119910119905+ (1 minus 119886) 119897

119905minus1 119910119905+ℎ|119905

= 119897119905

119910119905+1

= 119910119905+1|119905

(1 + 120576119905) 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 09999 119897 = 113736307 120590 = 01327

TO (A A N)119897119905= 120572119910119905+ (1 minus 120572) (119897

119905minus1+ 119887119905minus1) 119887119905= 120573 (119897

119905minus 119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905+ ℎ119887119905

119910119905+1

= 119910119905+1|119905

+ 120576119905 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 09999 120573 = 01665 119897 = 784219911 119887 = 93651674 120590 = 146462100

In the above formulae for recursive calculations and point forecasts 119897119905denotes the series level at time 119905 119887

119905denotes the

slope at time 119905 120572 and 120573 are smoothness index ℎ denotes the lag period

the development of technology the construction sector willbecome dependent on power and all kinds of oil productsincreasingly instead of coal which can improve the energyefficiency and control environmental pollution gradually

Wholesale and retail trade and hotels and catering ser-vices consume coal electricity heat gasoline diesel oil lique-fied petroleum gas natural gas and other varieties of energy

Influenced by price changes and economic development thissector has gone through booms and busts repeatedly From1980 to 1990 the sector had a rising trend due to adaptingto the market demand quickly from 1990 to 1997 with highprice the sector was in the integration stage from 1997 to1999 the sector was in a low price and low growth stagefrom 2000 to 2004 the sector was in a low price and high

8 Discrete Dynamics in Nature and Society

Table 3 Accuracy of forecast models on sector by ETS method

AFAFW IND CON TSP WRHC OS HC TOAIC 3107869 4694980 2392015 2724609 3085274 3287655 3715486 4747963AICc 3114536 4719980 2417015 2749609 3110274 3294321 3722153 4772963BIC 3128760 4736761 2433796 2766390 3127055 3308545 3736377 4789744ME minus01636 12971479 02783 53547 minus218557 455007 minus1054612 19036901RMSE 3244691 129103640 536530 1544569 3098943 4978256 15404774 146462092MAE 2556755 106357804 396210 1158266 2727594 3650402 10815817 118112234MPE minus40490 minus02898 minus10556 minus08693 minus105543 minus64377 minus19142 minus01609MAPE 157232 98982 77865 87576 271766 300246 96963 93149MASE 07907 07024 08702 07315 11268 09915 09609 07435

Table 4 Forecasting result of coal consumption on sector by ETS method

ETS TO AFAFW IND CON TSP WRHC OS HC2011 32882036 179379 31311760 64068 55260 155568 200086 9159172012 34593125 179379 33020350 64942 51787 160666 200086 9159172013 36304590 179379 34728930 65816 48532 165931 200086 9159172014 38016441 179379 36437520 66691 45481 171368 200086 9159172015 39728663 179379 38146110 67565 42623 176984 200086 9159172016 41441238 179379 39854690 68439 39944 182784 200086 9159172017 43154182 179379 41563280 69314 37433 188774 200086 9159172018 44867480 179379 43271870 70188 35080 194961 200086 9159172019 46581118 179379 44980450 71062 32875 201350 200086 9159172020 48295114 179379 46689040 71937 30809 207948 200086 915917Unit 10000 tons

growth stage domestic prices began to soar in 2005 andsince the American subprime crisis and the domestic policyadjustment in 2008 the sector appears to have had violentfluctuations However Figure 5 shows that the wholesale andretail trade hotels and catering services will have a growthtrend of coal consumption in the future Industry scale willcontinue to expand which is consistent with our countryrsquospolicy to develop the tertiary industry vigorously

Agriculture forestry animal husbandry fishery andwater conservancy are the foundation of the national econ-omy Coal consumption was in the second place before 1990but in 1993 2000 2005 and 2008 there was a great fall(Figure 5) in contrast coal consumption of wholesale andretail trade hotels and catering services and other sectorshad an obvious increase in these three years Considerablecontact with the change of the prices is guessed

6 Results and Discussion

Table 5 shows prediction results of BVAR models under thethree different prior distributions and the ETS models Theunit is 10000 tons in ETS models which is 10000 tons ofstandard coal in BVAR models So the results of ETS modelare multiplied by the converted coefficient of 07143 as tons ofstandard coal

The result of BVAR models under the normal-Wishartprior distribution (prior = 0) is the same as under the normal-flat prior distribution (prior = 1) Using the ETS model to

Table 5 Total coal consumption forecast

BVARPrior = 0 or 1

BVARPrior = 2

ETSaggregation

ETSforecast

2011 23803337 23803337 234876 2344752012 25268675 25403407 247099 2459212013 26815268 25766741 259324 2573662014 28449414 25711339 271551 2688112015 3017629 27462729 283782 2802562016 32001513 31472065 296014 2917012017 33931196 36070012 308250 3031462018 35972003 40119003 320489 3145912019 38131212 44491339 332729 3260372020 40416784 50482369 344972 337481Unit 10000 tons of standard coal

predict coal consumption of every sector and then addingthe seven values to get 344972 billion tons of standard coalin 2020 only based on the historical total coal consumption337481 billion tons of standard coal are obtained in 2020directly FromTable 5 in 2015 forecast results of fourmethodshave small differences But in 2020 the results of BVARmodelare larger than ETS model (Figure 6)

The reasonmay be that in BVARmodel the gross value ofindustrial output and the downstream industrial productionare influencing factors whose values are increasing over the

Discrete Dynamics in Nature and Society 9

20

25

30

35

40

45

50

55

2011 2012 2013 2014 2015 2016 2017 2018 2019 2020

BVAR 0 1BVAR 2

ETS AETS F

Coa

l con

sum

ptio

n (te

n th

ousa

nd to

ns o

f sta

ndar

d co

al)

times104

Figure 6 Total coal consumption forecast

years So they drive the rise in total coal consumption in thelong run However in ETS model total coal consumption isthe sum of coal consumption in seven sectors and the abovehas introduced that the coal consumption of some sectorshas showed a decreasing trend such as agriculture forestryanimal husbandry fishery and water conservancy transportstorage and post So the ETS model prediction results aresmaller

7 Conclusions

China is a big developing country in the process of industri-alization and urbanization and therefore coal consumptionwill remain high in the future for a long time Based on guar-anteeing the stability of the VAR system coal consumptiongross industrial output value and the downstream industrialproduction (crude steel cement thermal power) are selectedas variables to establish the BVAR (2) model At the sametime this paper analyses the changes of coal consumptionstructure based on sector division The forecast models areselected through comparing information criteria and forecastaccuracy comparison from thirty ETSmodels By forecastingthe total coal consumption and analyzing structural changesthis paper gets the following main conclusions

(1) Results of the impulse response function indicate thatthe increased output of the downstream industry willbecome a driving force of coal consumption In thefirst two periods the same response has been therewith the increase of the gross value of industrialoutput Later there will be a negative response

(2) The variance decomposition results show that thecontribution of gross value of industrial outputaccounted for 50of growth in coal consumption andhas long-term stable influence in the downstreamindustry the contribution rate of crude steel outputis greater than that of the cement output and thermalpower output

(3) In the next decade from a sector division point ofview industry and household consumption account

for the largest proportion of total coal consumptionthe proportion of industrial coal consumption willcontinue to rise but the proportion of householdcoal consumption will continue to decline Transportstorage and post sectors were the largest consumersof coal consumption in 1980 which will becomethe lowest in the future Construction sector despiteits lower reliance on coal has experienced a risein coal consumption in recent years The scales ofconsumption of wholesale and retail trade hotelsand catering services agriculture forestry animalhusbandry fishery and water conservancy sectorswere influenced by price fluctuations in the past fewdecades causing the fluctuation of coal consumptionBut in the next decade coal consumption of bothsectors will have a rising trend

(4) TheBVAR (2) forecastmodels predict that in 2015 coaldemand will reach 275ndash302 billion tons of standardcoal and in 2020 it will reach 404ndash505 billion tonsof standard coal in China That is to say as themomentumof rapid development of Chinarsquos economyin the future the coal consumption in 2020 is almostdouble that of in 2012 which is 246ndash253 billion tonsof standard coal However because the unrenewablecharacteristics of the coal and the pollution causedby burning are bound to draw the attention of thecountry and the residents how to improve the coalutilization efficiency and promote the use of cleanenergy will become a main research direction in thefuture

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This research is supported by the National Natural ScienceFoundation of China (no 71103115) the China Postdoc-toral Science Foundation surface of funded projects (no2012M510580) the Fundamental Research Funds for theShaanxi province soft science (no 2012KRM95) The authorswould like to thank the anonymous referees

References

[1] L Suganthi and A A Samuel ldquoEnergy models for demandforecastingmdasha reviewrdquo Renewable and Sustainable EnergyReviews vol 16 no 2 pp 1223ndash1240 2012

[2] V S Ediger and S Akar ldquoARIMA forecasting of primary energydemand by fuel in Turkeyrdquo Energy Policy vol 35 no 3 pp 1701ndash1708 2007

[3] S HMohr and GM Evans ldquoForecasting coal production until2100rdquo Fuel vol 88 no 11 pp 2059ndash2067 2009

[4] A Unler ldquoImprovement of energy demand forecasts usingswarm intelligence the case of Turkey with projections to 2025rdquoEnergy Policy vol 36 no 6 pp 1937ndash1944 2008

10 Discrete Dynamics in Nature and Society

[5] C-C Lee and Y-B Chiu ldquoModeling OECD energy demand aninternational panel smooth transition error-correction modelrdquoInternational Review of Economics amp Finance vol 25 pp 372ndash383 2013

[6] H Chan and S Lee ldquoForecasting the demand for energy inChinardquo Energy Journal vol 17 no 1 pp 19ndash30 1996

[7] C V Hirschhausen and M Andres ldquoLong-term electricitydemand in Chinamdashfrom quantitative to qualitative growthrdquoEnergy Policy vol 28 no 4 pp 231ndash241 2000

[8] S-W Yu and K-J Zhu ldquoA hybrid procedure for energy demandforecasting in Chinardquo Energy vol 37 no 1 pp 396ndash404 2012

[9] S W Yu Y-M Wei and K Wang ldquoChinas primary energydemands in 2020 predictions from an MPSO-RBF estimationmodelrdquo Energy Conversion andManagement vol 61 pp 59ndash662012

[10] M Zhang H L Mu G Li and Y D Ning ldquoForecasting thetransport energy demand based on PLSR method in ChinardquoEnergy vol 34 no 9 pp 1396ndash1400 2009

[11] P Crompton and Y RWu ldquoEnergy consumption in China pasttrends and future directionsrdquo Energy Economics vol 27 no 1pp 195ndash208 2005

[12] J Chai Z Y Zhang J L Fu J E Guo and S Y WangldquoIdentification and analysis of the structural mutations in oilprice systemrdquo Journal of Management Sciences vol 27 no 2 pp133ndash144 2014 (Chinese)

[13] BM Francis L Moseley and S O Iyare ldquoEnergy consumptionand projected growth in selected Caribbean countriesrdquo EnergyEconomics vol 29 no 6 pp 1224ndash1232 2007

[14] L M Saini ldquoPeak load forecasting using Bayesian regulariza-tion Resilient and adaptive backpropagation learning basedartificial neural networksrdquo Electric Power Systems Research vol78 no 7 pp 1302ndash1310 2008

[15] P Lauret E Fock R N Randrianarivony and J-F Manicom-Ramsamy ldquoBayesian neural network approach to short timeload forecastingrdquo Energy Conversion and Management vol 49no 5 pp 1156ndash1166 2008

[16] C A Sims ldquoMacroeconomics and realityrdquo Econometrica vol48 no 1 pp 1ndash48 1980

[17] T Zha ldquoA dynamic multivariate model for use in formulatingpolicyrdquo Economic Review vol 83 no 1 pp 16ndash29 1998

[18] R B Litterman ldquoA Bayesian procedure for forecasting withvector autoregressionsrdquo Department of Economics WorkingPaper Massachusetts Institute of Technology 1980

[19] R B Litterman ldquoForecasting with Bayesian vector autoregres-sions five years of experiencerdquo Journal of Business and EconomicStatistics vol 4 no 1 pp 25ndash38 1986

[20] T Doan R B Litterman and C A Sims ldquoForecastingand conditional projection using realistic prior distributionsrdquoEconometric Reviews vol 3 no 1 pp 1ndash100 1984

[21] C A Sims ldquoA nine-variable probabilistic macroeconomic fore-casting modelrdquo in Business Cycles Indicators and Forecasting JH Stock and M W Watson Eds University of Chicago PressChicago Ill USA 1992

[22] C A Sims and T A Zha ldquoBayesian methods for dynamicmultivariate modelsrdquo International Economic Review vol 39no 4 pp 949ndash968 1998

[23] R J Hyndman and Y Khandakar ldquoAutomatic time seriesforecasting the forecast package for Rrdquo Journal of StatisticalSoftware vol 27 no 3 pp 1ndash22 2008

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical Problems in Engineering

Hindawi Publishing Corporationhttpwwwhindawicom

Differential EquationsInternational Journal of

Volume 2014

Applied MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical PhysicsAdvances in

Complex AnalysisJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OptimizationJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Operations ResearchAdvances in

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Function Spaces

Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of Mathematics and Mathematical Sciences

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Algebra

Discrete Dynamics in Nature and Society

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Decision SciencesAdvances in

Discrete MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Stochastic AnalysisInternational Journal of

2 Discrete Dynamics in Nature and Society

cumulative production scenario best guess scenario) Theultimately recoverable resources (URR) estimates used in thescenarios ranged from700Gt to 1243GtUnler [4] proposed amodel to forecast the energy demand of Turkey until 2025 byusing PSO-based energy demand forecasting (PSOEDF) andmade a comparison with the ant colony optimization (ACO)energy demand estimation model The result showed thatthe former algorithm had better accuracy Lee and Chiu [5]applied a newly developed panel smooth transition regressionmodel with the error-correction term (PSECM) to estimatethe nonlinear relationships among energy consumption realincome and real energy prices for 24 OECD countries

For domestic energy consumption various approachesto forecasting have been used in the literature includingcointegration and error correction models [6] and demandequations [7] In 2002 Yu and Zhu [8] proposed animproved hybrid algorithm called PSO-GA (particle swarmoptimization-genetic algorithm) for energy demand forecast-ing in China with higher precision compared with singleoptimization methods such as GA PSO or ant colony opti-mization andmultiple linear regressions Results of this studyshow that Chinarsquos energy demandwill be 470 billion tons coalequivalent in 2015 In the same year they also proposed mix-encoding particle swarm optimization and radial basis func-tion (MPSO-RBF) network-based model to forecast Chinarsquosenergy consumption until 2020 based on GDP populationproportion of industry in GDP urbanization rate and shareof coal energy for the period from 1980 to 2009 [9] Zhanget al [10] forecast transport energy demand for 2010 2015and 2020 based on partial least square regression (PLSR)method under two scenarios by analyzing gross domesticproduct (GDP) urbanization rate passenger-turnover andfreight-turnover for the period of 1990ndash2006 Crompton andWu [11] applied the Bayesian vector autoregressive method-ology to forecast Chinarsquos energy consumption and to discusspotential implicationsThe results of this paper suggested thattotal energy consumption should increase to 2173 MtCE in2010 an annual growth rate of 38 which is slightly slowerthan the average rate in the past decade due to structuralchanges in the Chinese economy

Among them some of the latest techniques such asBayesian vector autoregression (BVAR) support vector re-gression ant colony and particle swarm optimizationmodelsare being used in energy demand analysis Chai et al [12]established a VARX (vector autoregressivemodel with exoge-nous variables) model of crude oil market structure to studythe effect of every variable on oil price by screening variableswith price supply demand of oil dollar index and Chinaoil net import as the endogenous variables and reservespeculative factors as the exogenous variablesThen based onthis VARX model and the Bayes theory a MSBVAR (Markovswitching bayes vector autoregression) model is establishedto identify and analyze the structural changing of oil pricesystem structure within the study period BVAR model andGranger-causality are applied to study growth in energydemand and the relationship between energy consumptionand real gross domestic product per capita in selected fewCaribbean countries [13] Bayesian neural network approachis used for short term electric load forecasting [14 15]

The BVAR model can avoid the rigid inclusionexclusionrestrictions of VAR models by allowing inclusion of manycoefficients while simultaneously controlling the extent ofmathematical expectation and standard deviation to whichthey can be influenced by the data This reduces the spuriouscorrelations captured by themodel thereby reducing forecasterror and improving forecasting performance

On the basis of the background and demand mentionedabove this paper attempts to analyze structural changes ofcoal consumption during the past 30 years and forecast futurecoal demand in China up to 2020 This paper includes sevenparts introduction describing the realistic background andthe academic background for research topic variables selec-tion and data processing screening the core effect factors fortotal coal demand methodology introducing VAR BVARand ETS models forecast of total coal consumption basedon VAR models applying the impulse response functionand variance decomposition to portray the dynamic corre-lations between coal consumption and economic variablesand forecasting total coal consumption by BVAR modelsstructural analysis of total coal consumption establishingETS models based on coal consumption of seven sectors toanalyze structural changes of coal consumption and forecasttotal coal consumption results and discussion comparingforecast results of total coal consumption by BVAR modelsand ETS models conclusions

2 Variables Selection and Data Processing

Many factors that affect future demand for coal such asdomestic and international price of coal the national policyof saving energy and reducing consumption efficiency ofresource utilization coal consumption of downstream indus-tries railway and highway capacity for coal transportationtechnology development urbanization industrial growthindustrial structure the supply and price of other alternativeenergy sources (such as oil natural gas hydropower nuclearpower etc) coal production cost (pollution control costscoal mine safety production costs exit cost etc) and con-sumer habits

Based upon the China Statistical Yearbook China EnergyStatistical Yearbook China Coal Industry Yearbook andWIND Database this paper has identified the followinginfluence factors of coal demand from 1985 to 2011 (1)unit of coal consumption is ten thousand tons of standardcoal (2) since most coal is consumed in industry andcontribution of the secondary industry to gross domesticproduct (GDP) fluctuates sharply gross industrial output (100million yuan) and industrial structure () are consideredimportant factors that affect coal consumption (3) electricpower steel building materials and chemical industry arefour major downstream industries that consume coal sooutput of thermal power (100 million kilowatt hour) crudesteel (10000 tons) and cement (10000 tons) are also factorsused to forecast coal consumption (4) effect of internationalprices and supply position of different substitute sources ofenergy on domestic coal usage and consumption such ascoal price (Newcastle in AustraliaKen blah FOB) oil price

Discrete Dynamics in Nature and Society 3

(Europe Brent Spot Price FOB) natural gas price (Louisianaspot prices) and producer price indices for domestic indus-trial products by sector (power coal and petroleum) (5)per capita GDP (yuan) and urbanization rate () data arecollected since the improvement of consumer capacity andchange of urban structure affect the energy consumption percapita in China

First of all in order to eliminate the heteroscedasticity ofthe economic time series data and to linearize its trend theabove variables are put into the natural logarithms Then toprevent false regression leading to an invalid conclusionADF(Augment Dikey-Fuller) unit root test and cointegration testare used to examine stationarity and long-run equilibriumrelationship of logarithmic time series respectively Finallyonly the following 5 groupsrsquo data meet the conditions ofthe VAR model coal consumption (COAL) gross value ofindustrial output (IND) output of cement industrial products(CEM) output of crude steel industrial products (STE) andoutput of thermal power industrial products (POW)

3 Methodology

31 VAR A vector autoregressive (VAR)model based on sta-tistical properties of the data is established without exertinga theory of a priori constraints for the data mechanism so itis an unstructured time series model not a structural econo-metric model with a theoretical foundation Specifically alleconomic variables are considered as endogenous variablesin VARmodel through multistage lag regression to estimatetheir relationships and for establishing forecast models

In 1980 Sims [16] introduced the VAR model and pro-moted its application for dynamic analysis of the economicsystem extensively VAR model is used to forecast inter-connected time series system and analyze dynamic shocksfrom stochastic disturbances to variables so as to explain theinfluence of various economic shocks on economic variables

An unrestricted vector autoregression (UVAR) modelcontaining 119899 time series variables and a lag length of 119901 hasthe following general form

119910119905= 119888 + 119860

1119910119905minus1

+ sdot sdot sdot + 119860119901119910119905minus119901

+ 120583119905

120583119905sim 119894119894119889119873 (0 120590

2) 119905 = 1 2 119899

(1)

where 119910119905= (1199101 1199102 119910

119899)1015840 is an (119899times1) vector of endogenous

variables 119888 is an (119899 times 1) vector of intercept terms 119860119894 119894 gt 1

are (119899 times 119899) coefficient matrices and 120583119905is an (119899 times 1) vector of

the independent normal random error vectorBased on the VAR model the dynamic relationship

between the variables can be analyzed by impulse responsefunction and variance decomposition

Impulse response function describes the effect on thecurrent and future variables from information shocks of arandom disturbance which equals a standard deviation andvividly depicts the path changes of the dynamic interactionbetween the variables and tests the intensity and durationof the impact of economic variables on coal consumptionThis function implies that decomposition does not have tocompletely depend on the order of the variables in the VAR

system so as to improve the stability and reliability of theestimation results

Variance decomposition method decomposes meansquare error (MSE) of the forecast system into contributionsof each variable facilitating examination of MSE decompo-sition of any endogenous variables and measurement of therelative importance of random disturbances for the variablesby comparing the variance contribution rate In this paperthrough the variance decomposition method the role of theeconomic variables fluctuation in coal consumption growthcan be determined

32 BVAR In (1) the vector 119888 contains 119899 intercept termsand each matrix 119860

119894contains 1198992 coefficients hence with the

increase of variables and lag 119899 + 1199011198992 coefficients must beestimated which increase exponentially with the numberof variables in the system A major problem in estimationof VAR models when 119901 is large is over-parameterizationwhere too many coefficients must be estimated relative tosample size This can lead to an overfitting problem the largenumber of coefficients in unrestricted VAR models tendsto fit the data unrealistically well while performing poorlyin out-of-sample forecasting due to the effects of spuriouscorrelations in the data set Over-fitting of the data candistort the long-run relationships between variables in themodel and inflate coefficient values on distant lags due to lowdegrees of freedom [17] In addition VAR model ignores thea priori information and presets the same importance for allestimated parameters which can lead to the wrong modelThere are many ways to overcome the over-fitting problemBy applying certain constraints on the parameters such asreducing the lag length or removing some of the variablesin an individual equation can solve the problem of lowerdegrees of freedom effectively But from the view of Bayesianthis means that forecasters believe that the probability ofremoving lags whose coefficient is zero is 100 But it isimpossible to know whether this constraint is established

An alternate approach is the BVAR method proposed byLitterman [18 19] Doan et al [20] Sims [21] and Sims andZha [22] The BVAR approach modifies the OLS estimatesof (1) by treating all coefficients as random variables aroundtheir Bayesian prior mean such that the model has theflexibility to impose these priors to varying degrees on thecoefficient estimates

33 ETS ETS technology appeared and was used in the1950s After decades of development it becamemature How-ever selection of ldquooptimumrdquo model from different modelsfor use began few years ago Hyndman and Khandakar[23] summed up exponential smoothing models and dividedthem into fifteen types For more details see Table 1

Error Trend and Seasonal of ETS model respectivelyrepresent error term trend term and season term in which(Trend Seasonal) group includes fifteen models listed inTable 1 Residual error term is classified as overlapping forms(addition forms) and multiplication forms On the assump-tion of 119910

119905= 120583119905+ 120576119905 it is an additive errors model On the

assumption of 119910119905= 120583119905(1 + 120576

119905) it is a multiplication errors

model Under the circumstances of consideration of different

4 Discrete Dynamics in Nature and Society

Table 1 Classification and summary of ETS models

Trend component Seasonal componentN (none) A (additive) M (multiplicative)

N (none) N N N A N MA (additive) A N A A A MAd (additive damped) Ad N Ad A Ad MM (multiplicative) M N M A M MMd (multiplicativedamped) Md N Md A Md M

error forms the above fifteen models can be extended tothirty types From the forecast results alone addition formor multiplication form hardly influence residual error termHowever for different sampled data various residual errorforms seem superior or inferior Because dividing by 0 isinvolved attention should be paid to select different additionor multiplication forms for residual error trend term orseason term When samples are all positive values usingresidual error multiplication form presents an advantageHowever when samples are zero value or negative valuemultiplication form model is not applicable

4 Forecast of Total Coal Consumption byBVAR Models

In order to analyze the dynamic relationship between coalconsumption and economic variables 5 d VARmodel is builtbased on (1) According to the lag length criteria (likelihoodratio final prediction error AIC SC and Hannan-Quinninformation criterion) VAR (2) is the most appropriatemodel VAR models are set up as follows

[[[[[

[

LNCOALLNINDLNCEMLNSTELNPOW

]]]]]

]

=

[[[[[

[

44768

minus200313

31506

minus24537

23387

]]]]]

]

+

[[[[[

[

06785 01126 minus01174 00534 04913

01305 05586 14616 minus04848 06649

minus06834 00470 10756 00198 06238

04434 minus02351 00606 08957 00600

minus02571 00139 01578 03482 09504

]]]]]

]

times

[[[[[

[

LNCOAL (minus1)LNIND (minus1)

LNCEM (minus1)

LNSTE (minus1)LNPOW (minus1)

]]]]]

]

00

05

10

15

00 05 10 15

minus05

minus10

minus15

minus05minus10minus15

Figure 1 Inverse roots of AR characteristic polynomial

+

[[[[[

[

minus03453 minus01248 minus00067 01699 minus02082

20973 minus04018 minus03793 minus05128 minus04853

03309 minus00134 minus07911 minus00454 03111

minus02974 00022 minus00554 minus02683 07020

minus00600 minus00030 minus01852 minus01937 00594

]]]]]

]

times

[[[[[

[

LNCOAL (minus2)LNIND (minus2)

LNCEM (minus2)

LNSTE (minus2)LNPOW (minus2)

]]]]]

]

+

[[[[[

[

1205761

1205762

1205763

1205764

1205765

]]]]]

]

(2)

The R-squared of the models are 09981 09951 09966and 09986 which indicate good simulation results At thesame time the reciprocals of all roots locate inside the unitcircle when calculating the AR characteristic polynomial(Figure 1) which shows that the established VAR (2) modelis stable That is to say when a variable changes in the model(ie to generate a shock) it leads to changes in other variablestoo But as time goes on the effect gradually disappears InFigure 1 the horizontal axis represents the real number axisthe vertical axis represents the imaginary numbers axis theunit circle is a circle of radius 1 and center at the origin thepoints in the unit circle represent roots of the characteristicequation on VAR (2) model

Based on the VAR model the dynamic relationshipbetween the variables can be analyzed by impulse responsefunction and variance decomposition

In Figure 2 the horizontal axis represents the lag periodsof shock from innovation (unit years) the vertical axisrepresents the response degree from the dependent variablesto the explaining variables the solid line is the calculatedvalue of the impulse response function The lag period is setto 15 years

Discrete Dynamics in Nature and Society 5

000

001

002

003

004

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

minus001

minus003

minus002

minus004

LNCOALLNINDLNCEM

LNSTELNPOW

Figure 2 Responses of LNCOAL to LNCOAL LNIND LNCEMLNSTE and LNPOW

According to Figure 2 the line graph shows that when theexplaining variables (coal consumption gross value of indus-trial output cement output crude steel industrial output andthermal power output) are shocked from innovation respec-tively how does the dependent variable (coal consumption)respond during the future 15 periods That is to say whenthese five factors change respectively we can forecast howmuch and how long will coal consumption respond

Firstly coal consumption volatility responds strongly toits own shock which is shown by the blue line and inthe first 15 periods it shows a fluctuating trend Early asa strong positive response in the second period it reachesthe peak at 00310 and then plunges to the lowest point atminus00145 in the eighth period From this point onwards theperiod between 9 and 15 experiences a gradual growth untilthe numbers become positive response again This showsthat coal consumption and its lagging values have a strongcorrelation Secondly after suffering a positive shock grossvalue of industrial output only brings positive response tocoal consumption in the second period and then it quicklydrops to a negative response to the lowest point of minus00379in the sixth period Afterwards it rises to zero slightly Thisindicates that the increase of the gross value of industrialoutput immediately leads to an increase in coal consumptionbut in the long term it reduces consumption The reasonsmay be that more output provides more capital for inter-nal restructuring and technological progress thus reducingcoal consumption costs and increasing profits Thirdly coalconsumption volatility responds modestly to cement outputshocks in the early stage From the fifth period to theeleventh period the response shows a large fluctuation fromnegative to positive Fourthly coal consumption volatilityhas increasing positive response to crude steel industrialoutput shocks gradually over time and in the seventh periodthat reaches the peak of 00219 Fifthly a positive shockon thermal power output brings positive response to coalconsumption which has weak intensity and stable tendencyThese results indicate that the development of downstream

0

20

40

60

80

100

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

LNCOALLNINDLNCEM

LNSTELNPOW

Figure 3 Factors variance decomposition on the effect of LNCOAL

industries drives an increase in demand for coal whose long-term relation is in the same direction

In Figure 3 the horizontal axis represents the lag periodsof impact from innovation (unit years) the vertical axisrepresents the contribution rate to coal consumption from theeconomic variables

As shown in Figure 3 in the future 15 periods coalconsumption has the greatest impact during the first fiveperiods with its own contribution rate above 40 Since thenrates in the rest of the period level off at about 22 Thecontribution rate of the gross value of industrial output to coalconsumption increases steeply from the third period and thenexceeds all other variables at the sixth period and remainsstable and high between 51 and 53Thermal power industryhad the highest contribution in the second period but afterthat its contribution rates decline to 3-4 Contribution ratesof crude steel industry have obvious rise from the thirdperiod and make the largest contribution in the downstreamindustries between 12ndash14 The contribution rate of cementindustry has an obvious rise in the sixth period after thatthe contribution rate remains in the 7-8 level Therefore inthe long run industrial output has the biggest contributionto the changes of coal consumption in a short time coalconsumption have the biggest contribution to own changes

The system of priors commonly used in the specificationof BVAR models includes conjugate prior distribution max-imum entropy prior distribution ML-II prior distributionand multilayer prior distribution Different BVAR priordistributions have different impacts on the prediction resultsThis paper uses the 119877 software to forecast coal consumptionbased on three kinds of prior distribution in the MSBVARpackage Command is run as follows Rgt fcastlt -szbvar (coal119901 = 2 lambda 0 = 06 lambda 1 = 01 lambda 3 = 2 lambda4 = 025 lambda 5 = 0 mu 5 = 0 mu 6 = 0 prior = 0)Thereare three different prior distributions 0 = Normal-Wishartprior 1 = Normal-flat prior and 2 = flat-flat prior (ie akintoMLE)The forecasting results from 2012 to 2020 are shownin Table 4

6 Discrete Dynamics in Nature and Society

6000

8000

10000

12000

14000

16000

18000

050000

100000150000200000250000300000350000400000450000500000

1980

1985

1990

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

2018

2019

2020

Industry coal consumptionHousehold coal consumption

Coa

l con

sum

ptio

n (1

0000

tons

)

Coa

l con

sum

ptio

n (1

0000

tons

)

Figure 4 Changes of industry and household coal consumption

5 Structural Analysis of TotalCoal Consumption

A univariate ETS (exponential smoothing) forecast model isused to predict coal consumption of each sector in 2020 inthe forecast package for R Error Trend and Seasonal of ETSmodel denote error term (A M Z) trend term (N A M Z)and season term (NAM Z) respectively inwhich the groupincludes thirty models

From the coal balance sheet in China Statistical YearbookandWINDDatabase coal consumption data of seven sectorsare selected for the period 1980ndash2010 The balance equa-tion for coal consumption Total coal consumption (TO) =Agriculture Forestry Animal Husbandry Fishery andWaterConservancy (AFAFW) + Industry (IND) + Construction(CON) + Transport Storage and Post (TSP) +Wholesale andRetail Trades Hotels and Catering Services (WRHC) +OtherSectors (OS) + Household Consumption (HC) Unit is 10000tons

The finalmodels (Table 2) are selected from thirtymodelsby comparing information criterion (AIC BIC AICc) andforecast precision (mean error root mean square error meanabsolute error maximum permissible errors mean absolutepercentage error mean absolute error square) (Table 3)Command is run as follows R gt fcast lt -forecast(est(x))

As can be seen from the Table 4 the total coal consump-tion is the sum of coal consumption in seven sectors whichis 39728663 billion tons in 2015 This value is similar tothe goal in the development of coal industry in the twelfthfive-year plan which is 39 billion tons in 2015 By 2020Chinarsquos total coal consumption will reach 483 billion tonswith an annual growth rate of 436 of which industry andhousehold consumption account for the largest proportionIn Figure 4 the industry coal consumption was 438 milliontons 811 million tons 1278 billion tons and 296 billiontons in 1980 1990 2000 and 2010 respectively and it ispredicted to reach 4669 billion tons in 2020 accountingfor 9667 of total coal consumption Therefore to reducecoal consumption in China industrial energy conservationand emissions reduction are the fastest and most effective

breakthrough points including the adjustment of industrialstructure and optimization of technology

However as the second largest main channel the propor-tion of household consumption is decreasing year by yearFigure 4 shows that household consumption was up to 167million tons in 1990 which accounts for 1583 of the totalcoal consumption By 2020 household consumption willbe 092 million tons accounting for 189 of the total coalconsumption In the field of household energy consumptionas the diversity of the type of energy and development oftechnology electricity and heating oil are more available andconvenient than coal for residents so less and less coal is usedin the household rapidly This is beneficial to environmentalprotection and nonrenewable resources saving

As shown in Figure 5 transport storage and post sectorswere the largest coal consumption sectors in 1980 but theybecame the lowest coal consumption sectors after thirty yearsin 2010 and their coal consumption will decline constantlyduring the next decade by predictionsThemain reason is thatcoal-fired steam locomotive engines have been replaced byinternal combustion and electric locomotives by the railwaysSo most of the raw coal has been replaced by fuel andelectricity at the same time Nowadays road transporta-tion mainly consumes gasoline and diesel oil waterwaystransportation mainly consumes fuel oil and diesel oil airtransport mainly consumes aviation kerosene Thereforealthough the development of transport storage and postsectors still consumes large amounts of energy the usage ofcoal is insignificant

As the pillar industry of national economy in our countryconstruction sector is also an energy-intensive industryCoal consumption increased from 556 to 71892 milliontons between 1980 and 2010 representing a relatively smallgrowth By forecasting the amount of coal consumption isstill increasing in the future (Figure 5) At the same time theconstruction sector does not have a high proportion of coalconsumption and accounted for only 023 in 2010 That isbecause there are various types of energies consumed by theconstruction industry including raw coal gasoline diesel oilfuel oil heat electricity and other petroleum products With

Discrete Dynamics in Nature and Society 7

300

600

900

1200

1500

1800

2100

2400

2700

1980

1985

1990

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

2018

2019

2020

AFAFWCONTSP

WRHCOS

Coa

l con

sum

ptio

n (1

0000

tons

)

Figure 5 Changes of other sectorsrsquo coal consumption

Table 2 Formulae for recursive calculations point forecasts and parametric estimated value on sector by ETS method

Variables Forms Models and parameters

AFAFW (M N N)119897119905= 120572119910119905+ (1 minus 120572) 119897

119905minus1 119910119905+ℎ|119905

= 119897119905

119910119905+1

= 119910119905+1|119905

(1 + 120576119905) 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 00001 119897 = 17937859 120590 = 01809

IND (A A N)

119897119905= 120572119910119905+ (1 minus 120572) (119897

119905minus1+ 119887119905minus1) 119887119905= 120573 (119897

119905minus 119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905+ ℎ119887119905

119910119905+1

= 119910119905+1|119905

+ 120576119905 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 09999 120573 = 02697119897 = 547821109 119887 = 93385498 120590 = 129103600

CON (A A N)119897119905= 120572119910119905+ (1 minus 120572) (119897

119905minus1+ 119887119905minus1) 119887119905= 120573 (119897

119905minus 119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905+ ℎ119887119905

119910119905+1

= 119910119905+1|119905

+ 120576119905 120576119905sim 119873119868119863 (0 120590

2) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 00009 120573 = 00009 119897 = 4483868 119887 = 87382 120590 = 536530

TSP (M M N)119897119905= 119886119910119905+ (1 minus 119886) 119897

119905minus1119887119905minus1

119887119905= 120573 (119897

119905119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905119887119905

ℎ

119910119905+1

= 119910119905+1|119905

(1 + 120576119905) 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 00001 120573 = 00001 119897 = 23050531 119887 = 09371 120590 = 01049

WRHC (M M N)119897119905= 119886119910119905+ (1 minus 119886) 119897

119905minus1119887119905minus1

119887119905= 120573 (119897

119905119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905119887119905

ℎ

119910119905+1

= 119910119905+1|119905

(1 + 120576119905) 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 00001 120573 = 00001 119897 = 7651436 119887 = 10328 120590 = 02562

OS (A N N)119897119905= 119886119910119905+ (1 minus 119886) 119897

119905minus1 119910119905+ℎ|119905

= 119897119905

119910119905+1

= 119910119905+1|119905

+ 120576119905 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 08917 119897 = 11487978 120590 = 4978256

HC (M N N)119897119905= 119886119910119905+ (1 minus 119886) 119897

119905minus1 119910119905+ℎ|119905

= 119897119905

119910119905+1

= 119910119905+1|119905

(1 + 120576119905) 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 09999 119897 = 113736307 120590 = 01327

TO (A A N)119897119905= 120572119910119905+ (1 minus 120572) (119897

119905minus1+ 119887119905minus1) 119887119905= 120573 (119897

119905minus 119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905+ ℎ119887119905

119910119905+1

= 119910119905+1|119905

+ 120576119905 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 09999 120573 = 01665 119897 = 784219911 119887 = 93651674 120590 = 146462100

In the above formulae for recursive calculations and point forecasts 119897119905denotes the series level at time 119905 119887

119905denotes the

slope at time 119905 120572 and 120573 are smoothness index ℎ denotes the lag period

the development of technology the construction sector willbecome dependent on power and all kinds of oil productsincreasingly instead of coal which can improve the energyefficiency and control environmental pollution gradually

Wholesale and retail trade and hotels and catering ser-vices consume coal electricity heat gasoline diesel oil lique-fied petroleum gas natural gas and other varieties of energy

Influenced by price changes and economic development thissector has gone through booms and busts repeatedly From1980 to 1990 the sector had a rising trend due to adaptingto the market demand quickly from 1990 to 1997 with highprice the sector was in the integration stage from 1997 to1999 the sector was in a low price and low growth stagefrom 2000 to 2004 the sector was in a low price and high

8 Discrete Dynamics in Nature and Society

Table 3 Accuracy of forecast models on sector by ETS method

AFAFW IND CON TSP WRHC OS HC TOAIC 3107869 4694980 2392015 2724609 3085274 3287655 3715486 4747963AICc 3114536 4719980 2417015 2749609 3110274 3294321 3722153 4772963BIC 3128760 4736761 2433796 2766390 3127055 3308545 3736377 4789744ME minus01636 12971479 02783 53547 minus218557 455007 minus1054612 19036901RMSE 3244691 129103640 536530 1544569 3098943 4978256 15404774 146462092MAE 2556755 106357804 396210 1158266 2727594 3650402 10815817 118112234MPE minus40490 minus02898 minus10556 minus08693 minus105543 minus64377 minus19142 minus01609MAPE 157232 98982 77865 87576 271766 300246 96963 93149MASE 07907 07024 08702 07315 11268 09915 09609 07435

Table 4 Forecasting result of coal consumption on sector by ETS method

ETS TO AFAFW IND CON TSP WRHC OS HC2011 32882036 179379 31311760 64068 55260 155568 200086 9159172012 34593125 179379 33020350 64942 51787 160666 200086 9159172013 36304590 179379 34728930 65816 48532 165931 200086 9159172014 38016441 179379 36437520 66691 45481 171368 200086 9159172015 39728663 179379 38146110 67565 42623 176984 200086 9159172016 41441238 179379 39854690 68439 39944 182784 200086 9159172017 43154182 179379 41563280 69314 37433 188774 200086 9159172018 44867480 179379 43271870 70188 35080 194961 200086 9159172019 46581118 179379 44980450 71062 32875 201350 200086 9159172020 48295114 179379 46689040 71937 30809 207948 200086 915917Unit 10000 tons

growth stage domestic prices began to soar in 2005 andsince the American subprime crisis and the domestic policyadjustment in 2008 the sector appears to have had violentfluctuations However Figure 5 shows that the wholesale andretail trade hotels and catering services will have a growthtrend of coal consumption in the future Industry scale willcontinue to expand which is consistent with our countryrsquospolicy to develop the tertiary industry vigorously

Agriculture forestry animal husbandry fishery andwater conservancy are the foundation of the national econ-omy Coal consumption was in the second place before 1990but in 1993 2000 2005 and 2008 there was a great fall(Figure 5) in contrast coal consumption of wholesale andretail trade hotels and catering services and other sectorshad an obvious increase in these three years Considerablecontact with the change of the prices is guessed

6 Results and Discussion

Table 5 shows prediction results of BVAR models under thethree different prior distributions and the ETS models Theunit is 10000 tons in ETS models which is 10000 tons ofstandard coal in BVAR models So the results of ETS modelare multiplied by the converted coefficient of 07143 as tons ofstandard coal

The result of BVAR models under the normal-Wishartprior distribution (prior = 0) is the same as under the normal-flat prior distribution (prior = 1) Using the ETS model to

Table 5 Total coal consumption forecast

BVARPrior = 0 or 1

BVARPrior = 2

ETSaggregation

ETSforecast

2011 23803337 23803337 234876 2344752012 25268675 25403407 247099 2459212013 26815268 25766741 259324 2573662014 28449414 25711339 271551 2688112015 3017629 27462729 283782 2802562016 32001513 31472065 296014 2917012017 33931196 36070012 308250 3031462018 35972003 40119003 320489 3145912019 38131212 44491339 332729 3260372020 40416784 50482369 344972 337481Unit 10000 tons of standard coal

predict coal consumption of every sector and then addingthe seven values to get 344972 billion tons of standard coalin 2020 only based on the historical total coal consumption337481 billion tons of standard coal are obtained in 2020directly FromTable 5 in 2015 forecast results of fourmethodshave small differences But in 2020 the results of BVARmodelare larger than ETS model (Figure 6)

The reasonmay be that in BVARmodel the gross value ofindustrial output and the downstream industrial productionare influencing factors whose values are increasing over the

Discrete Dynamics in Nature and Society 9

20

25

30

35

40

45

50

55

2011 2012 2013 2014 2015 2016 2017 2018 2019 2020

BVAR 0 1BVAR 2

ETS AETS F

Coa

l con

sum

ptio

n (te

n th

ousa

nd to

ns o

f sta

ndar

d co

al)

times104

Figure 6 Total coal consumption forecast

years So they drive the rise in total coal consumption in thelong run However in ETS model total coal consumption isthe sum of coal consumption in seven sectors and the abovehas introduced that the coal consumption of some sectorshas showed a decreasing trend such as agriculture forestryanimal husbandry fishery and water conservancy transportstorage and post So the ETS model prediction results aresmaller

7 Conclusions

China is a big developing country in the process of industri-alization and urbanization and therefore coal consumptionwill remain high in the future for a long time Based on guar-anteeing the stability of the VAR system coal consumptiongross industrial output value and the downstream industrialproduction (crude steel cement thermal power) are selectedas variables to establish the BVAR (2) model At the sametime this paper analyses the changes of coal consumptionstructure based on sector division The forecast models areselected through comparing information criteria and forecastaccuracy comparison from thirty ETSmodels By forecastingthe total coal consumption and analyzing structural changesthis paper gets the following main conclusions

(1) Results of the impulse response function indicate thatthe increased output of the downstream industry willbecome a driving force of coal consumption In thefirst two periods the same response has been therewith the increase of the gross value of industrialoutput Later there will be a negative response

(2) The variance decomposition results show that thecontribution of gross value of industrial outputaccounted for 50of growth in coal consumption andhas long-term stable influence in the downstreamindustry the contribution rate of crude steel outputis greater than that of the cement output and thermalpower output

(3) In the next decade from a sector division point ofview industry and household consumption account

for the largest proportion of total coal consumptionthe proportion of industrial coal consumption willcontinue to rise but the proportion of householdcoal consumption will continue to decline Transportstorage and post sectors were the largest consumersof coal consumption in 1980 which will becomethe lowest in the future Construction sector despiteits lower reliance on coal has experienced a risein coal consumption in recent years The scales ofconsumption of wholesale and retail trade hotelsand catering services agriculture forestry animalhusbandry fishery and water conservancy sectorswere influenced by price fluctuations in the past fewdecades causing the fluctuation of coal consumptionBut in the next decade coal consumption of bothsectors will have a rising trend

(4) TheBVAR (2) forecastmodels predict that in 2015 coaldemand will reach 275ndash302 billion tons of standardcoal and in 2020 it will reach 404ndash505 billion tonsof standard coal in China That is to say as themomentumof rapid development of Chinarsquos economyin the future the coal consumption in 2020 is almostdouble that of in 2012 which is 246ndash253 billion tonsof standard coal However because the unrenewablecharacteristics of the coal and the pollution causedby burning are bound to draw the attention of thecountry and the residents how to improve the coalutilization efficiency and promote the use of cleanenergy will become a main research direction in thefuture

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This research is supported by the National Natural ScienceFoundation of China (no 71103115) the China Postdoc-toral Science Foundation surface of funded projects (no2012M510580) the Fundamental Research Funds for theShaanxi province soft science (no 2012KRM95) The authorswould like to thank the anonymous referees

References

[1] L Suganthi and A A Samuel ldquoEnergy models for demandforecastingmdasha reviewrdquo Renewable and Sustainable EnergyReviews vol 16 no 2 pp 1223ndash1240 2012

[2] V S Ediger and S Akar ldquoARIMA forecasting of primary energydemand by fuel in Turkeyrdquo Energy Policy vol 35 no 3 pp 1701ndash1708 2007

[3] S HMohr and GM Evans ldquoForecasting coal production until2100rdquo Fuel vol 88 no 11 pp 2059ndash2067 2009

[4] A Unler ldquoImprovement of energy demand forecasts usingswarm intelligence the case of Turkey with projections to 2025rdquoEnergy Policy vol 36 no 6 pp 1937ndash1944 2008

10 Discrete Dynamics in Nature and Society

[5] C-C Lee and Y-B Chiu ldquoModeling OECD energy demand aninternational panel smooth transition error-correction modelrdquoInternational Review of Economics amp Finance vol 25 pp 372ndash383 2013

[6] H Chan and S Lee ldquoForecasting the demand for energy inChinardquo Energy Journal vol 17 no 1 pp 19ndash30 1996

[7] C V Hirschhausen and M Andres ldquoLong-term electricitydemand in Chinamdashfrom quantitative to qualitative growthrdquoEnergy Policy vol 28 no 4 pp 231ndash241 2000

[8] S-W Yu and K-J Zhu ldquoA hybrid procedure for energy demandforecasting in Chinardquo Energy vol 37 no 1 pp 396ndash404 2012

[9] S W Yu Y-M Wei and K Wang ldquoChinas primary energydemands in 2020 predictions from an MPSO-RBF estimationmodelrdquo Energy Conversion andManagement vol 61 pp 59ndash662012

[10] M Zhang H L Mu G Li and Y D Ning ldquoForecasting thetransport energy demand based on PLSR method in ChinardquoEnergy vol 34 no 9 pp 1396ndash1400 2009

[11] P Crompton and Y RWu ldquoEnergy consumption in China pasttrends and future directionsrdquo Energy Economics vol 27 no 1pp 195ndash208 2005

[12] J Chai Z Y Zhang J L Fu J E Guo and S Y WangldquoIdentification and analysis of the structural mutations in oilprice systemrdquo Journal of Management Sciences vol 27 no 2 pp133ndash144 2014 (Chinese)

[13] BM Francis L Moseley and S O Iyare ldquoEnergy consumptionand projected growth in selected Caribbean countriesrdquo EnergyEconomics vol 29 no 6 pp 1224ndash1232 2007

[14] L M Saini ldquoPeak load forecasting using Bayesian regulariza-tion Resilient and adaptive backpropagation learning basedartificial neural networksrdquo Electric Power Systems Research vol78 no 7 pp 1302ndash1310 2008

[15] P Lauret E Fock R N Randrianarivony and J-F Manicom-Ramsamy ldquoBayesian neural network approach to short timeload forecastingrdquo Energy Conversion and Management vol 49no 5 pp 1156ndash1166 2008

[16] C A Sims ldquoMacroeconomics and realityrdquo Econometrica vol48 no 1 pp 1ndash48 1980

[17] T Zha ldquoA dynamic multivariate model for use in formulatingpolicyrdquo Economic Review vol 83 no 1 pp 16ndash29 1998

[18] R B Litterman ldquoA Bayesian procedure for forecasting withvector autoregressionsrdquo Department of Economics WorkingPaper Massachusetts Institute of Technology 1980

[19] R B Litterman ldquoForecasting with Bayesian vector autoregres-sions five years of experiencerdquo Journal of Business and EconomicStatistics vol 4 no 1 pp 25ndash38 1986

[20] T Doan R B Litterman and C A Sims ldquoForecastingand conditional projection using realistic prior distributionsrdquoEconometric Reviews vol 3 no 1 pp 1ndash100 1984

[21] C A Sims ldquoA nine-variable probabilistic macroeconomic fore-casting modelrdquo in Business Cycles Indicators and Forecasting JH Stock and M W Watson Eds University of Chicago PressChicago Ill USA 1992

[22] C A Sims and T A Zha ldquoBayesian methods for dynamicmultivariate modelsrdquo International Economic Review vol 39no 4 pp 949ndash968 1998

[23] R J Hyndman and Y Khandakar ldquoAutomatic time seriesforecasting the forecast package for Rrdquo Journal of StatisticalSoftware vol 27 no 3 pp 1ndash22 2008

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical Problems in Engineering

Hindawi Publishing Corporationhttpwwwhindawicom

Differential EquationsInternational Journal of

Volume 2014

Applied MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical PhysicsAdvances in

Complex AnalysisJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OptimizationJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Operations ResearchAdvances in

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Function Spaces

Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of Mathematics and Mathematical Sciences

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Algebra

Discrete Dynamics in Nature and Society

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Decision SciencesAdvances in

Discrete MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Stochastic AnalysisInternational Journal of

Discrete Dynamics in Nature and Society 3

(Europe Brent Spot Price FOB) natural gas price (Louisianaspot prices) and producer price indices for domestic indus-trial products by sector (power coal and petroleum) (5)per capita GDP (yuan) and urbanization rate () data arecollected since the improvement of consumer capacity andchange of urban structure affect the energy consumption percapita in China

First of all in order to eliminate the heteroscedasticity ofthe economic time series data and to linearize its trend theabove variables are put into the natural logarithms Then toprevent false regression leading to an invalid conclusionADF(Augment Dikey-Fuller) unit root test and cointegration testare used to examine stationarity and long-run equilibriumrelationship of logarithmic time series respectively Finallyonly the following 5 groupsrsquo data meet the conditions ofthe VAR model coal consumption (COAL) gross value ofindustrial output (IND) output of cement industrial products(CEM) output of crude steel industrial products (STE) andoutput of thermal power industrial products (POW)

3 Methodology

31 VAR A vector autoregressive (VAR)model based on sta-tistical properties of the data is established without exertinga theory of a priori constraints for the data mechanism so itis an unstructured time series model not a structural econo-metric model with a theoretical foundation Specifically alleconomic variables are considered as endogenous variablesin VARmodel through multistage lag regression to estimatetheir relationships and for establishing forecast models

In 1980 Sims [16] introduced the VAR model and pro-moted its application for dynamic analysis of the economicsystem extensively VAR model is used to forecast inter-connected time series system and analyze dynamic shocksfrom stochastic disturbances to variables so as to explain theinfluence of various economic shocks on economic variables

An unrestricted vector autoregression (UVAR) modelcontaining 119899 time series variables and a lag length of 119901 hasthe following general form

119910119905= 119888 + 119860

1119910119905minus1

+ sdot sdot sdot + 119860119901119910119905minus119901

+ 120583119905

120583119905sim 119894119894119889119873 (0 120590

2) 119905 = 1 2 119899

(1)

where 119910119905= (1199101 1199102 119910

119899)1015840 is an (119899times1) vector of endogenous

variables 119888 is an (119899 times 1) vector of intercept terms 119860119894 119894 gt 1

are (119899 times 119899) coefficient matrices and 120583119905is an (119899 times 1) vector of

the independent normal random error vectorBased on the VAR model the dynamic relationship

between the variables can be analyzed by impulse responsefunction and variance decomposition

Impulse response function describes the effect on thecurrent and future variables from information shocks of arandom disturbance which equals a standard deviation andvividly depicts the path changes of the dynamic interactionbetween the variables and tests the intensity and durationof the impact of economic variables on coal consumptionThis function implies that decomposition does not have tocompletely depend on the order of the variables in the VAR

system so as to improve the stability and reliability of theestimation results

Variance decomposition method decomposes meansquare error (MSE) of the forecast system into contributionsof each variable facilitating examination of MSE decompo-sition of any endogenous variables and measurement of therelative importance of random disturbances for the variablesby comparing the variance contribution rate In this paperthrough the variance decomposition method the role of theeconomic variables fluctuation in coal consumption growthcan be determined

32 BVAR In (1) the vector 119888 contains 119899 intercept termsand each matrix 119860

119894contains 1198992 coefficients hence with the

increase of variables and lag 119899 + 1199011198992 coefficients must beestimated which increase exponentially with the numberof variables in the system A major problem in estimationof VAR models when 119901 is large is over-parameterizationwhere too many coefficients must be estimated relative tosample size This can lead to an overfitting problem the largenumber of coefficients in unrestricted VAR models tendsto fit the data unrealistically well while performing poorlyin out-of-sample forecasting due to the effects of spuriouscorrelations in the data set Over-fitting of the data candistort the long-run relationships between variables in themodel and inflate coefficient values on distant lags due to lowdegrees of freedom [17] In addition VAR model ignores thea priori information and presets the same importance for allestimated parameters which can lead to the wrong modelThere are many ways to overcome the over-fitting problemBy applying certain constraints on the parameters such asreducing the lag length or removing some of the variablesin an individual equation can solve the problem of lowerdegrees of freedom effectively But from the view of Bayesianthis means that forecasters believe that the probability ofremoving lags whose coefficient is zero is 100 But it isimpossible to know whether this constraint is established

An alternate approach is the BVAR method proposed byLitterman [18 19] Doan et al [20] Sims [21] and Sims andZha [22] The BVAR approach modifies the OLS estimatesof (1) by treating all coefficients as random variables aroundtheir Bayesian prior mean such that the model has theflexibility to impose these priors to varying degrees on thecoefficient estimates

33 ETS ETS technology appeared and was used in the1950s After decades of development it becamemature How-ever selection of ldquooptimumrdquo model from different modelsfor use began few years ago Hyndman and Khandakar[23] summed up exponential smoothing models and dividedthem into fifteen types For more details see Table 1

Error Trend and Seasonal of ETS model respectivelyrepresent error term trend term and season term in which(Trend Seasonal) group includes fifteen models listed inTable 1 Residual error term is classified as overlapping forms(addition forms) and multiplication forms On the assump-tion of 119910

119905= 120583119905+ 120576119905 it is an additive errors model On the

assumption of 119910119905= 120583119905(1 + 120576

119905) it is a multiplication errors

model Under the circumstances of consideration of different

4 Discrete Dynamics in Nature and Society

Table 1 Classification and summary of ETS models

Trend component Seasonal componentN (none) A (additive) M (multiplicative)

N (none) N N N A N MA (additive) A N A A A MAd (additive damped) Ad N Ad A Ad MM (multiplicative) M N M A M MMd (multiplicativedamped) Md N Md A Md M

error forms the above fifteen models can be extended tothirty types From the forecast results alone addition formor multiplication form hardly influence residual error termHowever for different sampled data various residual errorforms seem superior or inferior Because dividing by 0 isinvolved attention should be paid to select different additionor multiplication forms for residual error trend term orseason term When samples are all positive values usingresidual error multiplication form presents an advantageHowever when samples are zero value or negative valuemultiplication form model is not applicable

4 Forecast of Total Coal Consumption byBVAR Models

In order to analyze the dynamic relationship between coalconsumption and economic variables 5 d VARmodel is builtbased on (1) According to the lag length criteria (likelihoodratio final prediction error AIC SC and Hannan-Quinninformation criterion) VAR (2) is the most appropriatemodel VAR models are set up as follows

[[[[[

[

LNCOALLNINDLNCEMLNSTELNPOW

]]]]]

]

=

[[[[[

[

44768

minus200313

31506

minus24537

23387

]]]]]

]

+

[[[[[

[

06785 01126 minus01174 00534 04913

01305 05586 14616 minus04848 06649

minus06834 00470 10756 00198 06238

04434 minus02351 00606 08957 00600

minus02571 00139 01578 03482 09504

]]]]]

]

times

[[[[[

[

LNCOAL (minus1)LNIND (minus1)

LNCEM (minus1)

LNSTE (minus1)LNPOW (minus1)

]]]]]

]

00

05

10

15

00 05 10 15

minus05

minus10

minus15

minus05minus10minus15

Figure 1 Inverse roots of AR characteristic polynomial

+

[[[[[

[

minus03453 minus01248 minus00067 01699 minus02082

20973 minus04018 minus03793 minus05128 minus04853

03309 minus00134 minus07911 minus00454 03111

minus02974 00022 minus00554 minus02683 07020

minus00600 minus00030 minus01852 minus01937 00594

]]]]]

]

times

[[[[[

[

LNCOAL (minus2)LNIND (minus2)

LNCEM (minus2)

LNSTE (minus2)LNPOW (minus2)

]]]]]

]

+

[[[[[

[

1205761

1205762

1205763

1205764

1205765

]]]]]

]

(2)

The R-squared of the models are 09981 09951 09966and 09986 which indicate good simulation results At thesame time the reciprocals of all roots locate inside the unitcircle when calculating the AR characteristic polynomial(Figure 1) which shows that the established VAR (2) modelis stable That is to say when a variable changes in the model(ie to generate a shock) it leads to changes in other variablestoo But as time goes on the effect gradually disappears InFigure 1 the horizontal axis represents the real number axisthe vertical axis represents the imaginary numbers axis theunit circle is a circle of radius 1 and center at the origin thepoints in the unit circle represent roots of the characteristicequation on VAR (2) model

Based on the VAR model the dynamic relationshipbetween the variables can be analyzed by impulse responsefunction and variance decomposition

In Figure 2 the horizontal axis represents the lag periodsof shock from innovation (unit years) the vertical axisrepresents the response degree from the dependent variablesto the explaining variables the solid line is the calculatedvalue of the impulse response function The lag period is setto 15 years

Discrete Dynamics in Nature and Society 5

000

001

002

003

004

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

minus001

minus003

minus002

minus004

LNCOALLNINDLNCEM

LNSTELNPOW

Figure 2 Responses of LNCOAL to LNCOAL LNIND LNCEMLNSTE and LNPOW

According to Figure 2 the line graph shows that when theexplaining variables (coal consumption gross value of indus-trial output cement output crude steel industrial output andthermal power output) are shocked from innovation respec-tively how does the dependent variable (coal consumption)respond during the future 15 periods That is to say whenthese five factors change respectively we can forecast howmuch and how long will coal consumption respond

Firstly coal consumption volatility responds strongly toits own shock which is shown by the blue line and inthe first 15 periods it shows a fluctuating trend Early asa strong positive response in the second period it reachesthe peak at 00310 and then plunges to the lowest point atminus00145 in the eighth period From this point onwards theperiod between 9 and 15 experiences a gradual growth untilthe numbers become positive response again This showsthat coal consumption and its lagging values have a strongcorrelation Secondly after suffering a positive shock grossvalue of industrial output only brings positive response tocoal consumption in the second period and then it quicklydrops to a negative response to the lowest point of minus00379in the sixth period Afterwards it rises to zero slightly Thisindicates that the increase of the gross value of industrialoutput immediately leads to an increase in coal consumptionbut in the long term it reduces consumption The reasonsmay be that more output provides more capital for inter-nal restructuring and technological progress thus reducingcoal consumption costs and increasing profits Thirdly coalconsumption volatility responds modestly to cement outputshocks in the early stage From the fifth period to theeleventh period the response shows a large fluctuation fromnegative to positive Fourthly coal consumption volatilityhas increasing positive response to crude steel industrialoutput shocks gradually over time and in the seventh periodthat reaches the peak of 00219 Fifthly a positive shockon thermal power output brings positive response to coalconsumption which has weak intensity and stable tendencyThese results indicate that the development of downstream

0

20

40

60

80

100

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

LNCOALLNINDLNCEM

LNSTELNPOW

Figure 3 Factors variance decomposition on the effect of LNCOAL

industries drives an increase in demand for coal whose long-term relation is in the same direction

In Figure 3 the horizontal axis represents the lag periodsof impact from innovation (unit years) the vertical axisrepresents the contribution rate to coal consumption from theeconomic variables

As shown in Figure 3 in the future 15 periods coalconsumption has the greatest impact during the first fiveperiods with its own contribution rate above 40 Since thenrates in the rest of the period level off at about 22 Thecontribution rate of the gross value of industrial output to coalconsumption increases steeply from the third period and thenexceeds all other variables at the sixth period and remainsstable and high between 51 and 53Thermal power industryhad the highest contribution in the second period but afterthat its contribution rates decline to 3-4 Contribution ratesof crude steel industry have obvious rise from the thirdperiod and make the largest contribution in the downstreamindustries between 12ndash14 The contribution rate of cementindustry has an obvious rise in the sixth period after thatthe contribution rate remains in the 7-8 level Therefore inthe long run industrial output has the biggest contributionto the changes of coal consumption in a short time coalconsumption have the biggest contribution to own changes

The system of priors commonly used in the specificationof BVAR models includes conjugate prior distribution max-imum entropy prior distribution ML-II prior distributionand multilayer prior distribution Different BVAR priordistributions have different impacts on the prediction resultsThis paper uses the 119877 software to forecast coal consumptionbased on three kinds of prior distribution in the MSBVARpackage Command is run as follows Rgt fcastlt -szbvar (coal119901 = 2 lambda 0 = 06 lambda 1 = 01 lambda 3 = 2 lambda4 = 025 lambda 5 = 0 mu 5 = 0 mu 6 = 0 prior = 0)Thereare three different prior distributions 0 = Normal-Wishartprior 1 = Normal-flat prior and 2 = flat-flat prior (ie akintoMLE)The forecasting results from 2012 to 2020 are shownin Table 4

6 Discrete Dynamics in Nature and Society

6000

8000

10000

12000

14000

16000

18000

050000

100000150000200000250000300000350000400000450000500000

1980

1985

1990

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

2018

2019

2020

Industry coal consumptionHousehold coal consumption

Coa

l con

sum

ptio

n (1

0000

tons

)

Coa

l con

sum

ptio

n (1

0000

tons

)

Figure 4 Changes of industry and household coal consumption

5 Structural Analysis of TotalCoal Consumption

A univariate ETS (exponential smoothing) forecast model isused to predict coal consumption of each sector in 2020 inthe forecast package for R Error Trend and Seasonal of ETSmodel denote error term (A M Z) trend term (N A M Z)and season term (NAM Z) respectively inwhich the groupincludes thirty models

From the coal balance sheet in China Statistical YearbookandWINDDatabase coal consumption data of seven sectorsare selected for the period 1980ndash2010 The balance equa-tion for coal consumption Total coal consumption (TO) =Agriculture Forestry Animal Husbandry Fishery andWaterConservancy (AFAFW) + Industry (IND) + Construction(CON) + Transport Storage and Post (TSP) +Wholesale andRetail Trades Hotels and Catering Services (WRHC) +OtherSectors (OS) + Household Consumption (HC) Unit is 10000tons

The finalmodels (Table 2) are selected from thirtymodelsby comparing information criterion (AIC BIC AICc) andforecast precision (mean error root mean square error meanabsolute error maximum permissible errors mean absolutepercentage error mean absolute error square) (Table 3)Command is run as follows R gt fcast lt -forecast(est(x))

As can be seen from the Table 4 the total coal consump-tion is the sum of coal consumption in seven sectors whichis 39728663 billion tons in 2015 This value is similar tothe goal in the development of coal industry in the twelfthfive-year plan which is 39 billion tons in 2015 By 2020Chinarsquos total coal consumption will reach 483 billion tonswith an annual growth rate of 436 of which industry andhousehold consumption account for the largest proportionIn Figure 4 the industry coal consumption was 438 milliontons 811 million tons 1278 billion tons and 296 billiontons in 1980 1990 2000 and 2010 respectively and it ispredicted to reach 4669 billion tons in 2020 accountingfor 9667 of total coal consumption Therefore to reducecoal consumption in China industrial energy conservationand emissions reduction are the fastest and most effective

breakthrough points including the adjustment of industrialstructure and optimization of technology

However as the second largest main channel the propor-tion of household consumption is decreasing year by yearFigure 4 shows that household consumption was up to 167million tons in 1990 which accounts for 1583 of the totalcoal consumption By 2020 household consumption willbe 092 million tons accounting for 189 of the total coalconsumption In the field of household energy consumptionas the diversity of the type of energy and development oftechnology electricity and heating oil are more available andconvenient than coal for residents so less and less coal is usedin the household rapidly This is beneficial to environmentalprotection and nonrenewable resources saving

As shown in Figure 5 transport storage and post sectorswere the largest coal consumption sectors in 1980 but theybecame the lowest coal consumption sectors after thirty yearsin 2010 and their coal consumption will decline constantlyduring the next decade by predictionsThemain reason is thatcoal-fired steam locomotive engines have been replaced byinternal combustion and electric locomotives by the railwaysSo most of the raw coal has been replaced by fuel andelectricity at the same time Nowadays road transporta-tion mainly consumes gasoline and diesel oil waterwaystransportation mainly consumes fuel oil and diesel oil airtransport mainly consumes aviation kerosene Thereforealthough the development of transport storage and postsectors still consumes large amounts of energy the usage ofcoal is insignificant

As the pillar industry of national economy in our countryconstruction sector is also an energy-intensive industryCoal consumption increased from 556 to 71892 milliontons between 1980 and 2010 representing a relatively smallgrowth By forecasting the amount of coal consumption isstill increasing in the future (Figure 5) At the same time theconstruction sector does not have a high proportion of coalconsumption and accounted for only 023 in 2010 That isbecause there are various types of energies consumed by theconstruction industry including raw coal gasoline diesel oilfuel oil heat electricity and other petroleum products With

Discrete Dynamics in Nature and Society 7

300

600

900

1200

1500

1800

2100

2400

2700

1980

1985

1990

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

2018

2019

2020

AFAFWCONTSP

WRHCOS

Coa

l con

sum

ptio

n (1

0000

tons

)

Figure 5 Changes of other sectorsrsquo coal consumption

Table 2 Formulae for recursive calculations point forecasts and parametric estimated value on sector by ETS method

Variables Forms Models and parameters

AFAFW (M N N)119897119905= 120572119910119905+ (1 minus 120572) 119897

119905minus1 119910119905+ℎ|119905

= 119897119905

119910119905+1

= 119910119905+1|119905

(1 + 120576119905) 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 00001 119897 = 17937859 120590 = 01809

IND (A A N)

119897119905= 120572119910119905+ (1 minus 120572) (119897

119905minus1+ 119887119905minus1) 119887119905= 120573 (119897

119905minus 119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905+ ℎ119887119905

119910119905+1

= 119910119905+1|119905

+ 120576119905 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 09999 120573 = 02697119897 = 547821109 119887 = 93385498 120590 = 129103600

CON (A A N)119897119905= 120572119910119905+ (1 minus 120572) (119897

119905minus1+ 119887119905minus1) 119887119905= 120573 (119897

119905minus 119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905+ ℎ119887119905

119910119905+1

= 119910119905+1|119905

+ 120576119905 120576119905sim 119873119868119863 (0 120590

2) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 00009 120573 = 00009 119897 = 4483868 119887 = 87382 120590 = 536530

TSP (M M N)119897119905= 119886119910119905+ (1 minus 119886) 119897

119905minus1119887119905minus1

119887119905= 120573 (119897

119905119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905119887119905

ℎ

119910119905+1

= 119910119905+1|119905

(1 + 120576119905) 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 00001 120573 = 00001 119897 = 23050531 119887 = 09371 120590 = 01049

WRHC (M M N)119897119905= 119886119910119905+ (1 minus 119886) 119897

119905minus1119887119905minus1

119887119905= 120573 (119897

119905119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905119887119905

ℎ

119910119905+1

= 119910119905+1|119905

(1 + 120576119905) 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 00001 120573 = 00001 119897 = 7651436 119887 = 10328 120590 = 02562

OS (A N N)119897119905= 119886119910119905+ (1 minus 119886) 119897

119905minus1 119910119905+ℎ|119905

= 119897119905

119910119905+1

= 119910119905+1|119905

+ 120576119905 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 08917 119897 = 11487978 120590 = 4978256

HC (M N N)119897119905= 119886119910119905+ (1 minus 119886) 119897

119905minus1 119910119905+ℎ|119905

= 119897119905

119910119905+1

= 119910119905+1|119905

(1 + 120576119905) 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 09999 119897 = 113736307 120590 = 01327

TO (A A N)119897119905= 120572119910119905+ (1 minus 120572) (119897

119905minus1+ 119887119905minus1) 119887119905= 120573 (119897

119905minus 119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905+ ℎ119887119905

119910119905+1

= 119910119905+1|119905

+ 120576119905 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 09999 120573 = 01665 119897 = 784219911 119887 = 93651674 120590 = 146462100

In the above formulae for recursive calculations and point forecasts 119897119905denotes the series level at time 119905 119887

119905denotes the

slope at time 119905 120572 and 120573 are smoothness index ℎ denotes the lag period

the development of technology the construction sector willbecome dependent on power and all kinds of oil productsincreasingly instead of coal which can improve the energyefficiency and control environmental pollution gradually

Wholesale and retail trade and hotels and catering ser-vices consume coal electricity heat gasoline diesel oil lique-fied petroleum gas natural gas and other varieties of energy

Influenced by price changes and economic development thissector has gone through booms and busts repeatedly From1980 to 1990 the sector had a rising trend due to adaptingto the market demand quickly from 1990 to 1997 with highprice the sector was in the integration stage from 1997 to1999 the sector was in a low price and low growth stagefrom 2000 to 2004 the sector was in a low price and high

8 Discrete Dynamics in Nature and Society

Table 3 Accuracy of forecast models on sector by ETS method

AFAFW IND CON TSP WRHC OS HC TOAIC 3107869 4694980 2392015 2724609 3085274 3287655 3715486 4747963AICc 3114536 4719980 2417015 2749609 3110274 3294321 3722153 4772963BIC 3128760 4736761 2433796 2766390 3127055 3308545 3736377 4789744ME minus01636 12971479 02783 53547 minus218557 455007 minus1054612 19036901RMSE 3244691 129103640 536530 1544569 3098943 4978256 15404774 146462092MAE 2556755 106357804 396210 1158266 2727594 3650402 10815817 118112234MPE minus40490 minus02898 minus10556 minus08693 minus105543 minus64377 minus19142 minus01609MAPE 157232 98982 77865 87576 271766 300246 96963 93149MASE 07907 07024 08702 07315 11268 09915 09609 07435

Table 4 Forecasting result of coal consumption on sector by ETS method

ETS TO AFAFW IND CON TSP WRHC OS HC2011 32882036 179379 31311760 64068 55260 155568 200086 9159172012 34593125 179379 33020350 64942 51787 160666 200086 9159172013 36304590 179379 34728930 65816 48532 165931 200086 9159172014 38016441 179379 36437520 66691 45481 171368 200086 9159172015 39728663 179379 38146110 67565 42623 176984 200086 9159172016 41441238 179379 39854690 68439 39944 182784 200086 9159172017 43154182 179379 41563280 69314 37433 188774 200086 9159172018 44867480 179379 43271870 70188 35080 194961 200086 9159172019 46581118 179379 44980450 71062 32875 201350 200086 9159172020 48295114 179379 46689040 71937 30809 207948 200086 915917Unit 10000 tons

growth stage domestic prices began to soar in 2005 andsince the American subprime crisis and the domestic policyadjustment in 2008 the sector appears to have had violentfluctuations However Figure 5 shows that the wholesale andretail trade hotels and catering services will have a growthtrend of coal consumption in the future Industry scale willcontinue to expand which is consistent with our countryrsquospolicy to develop the tertiary industry vigorously

Agriculture forestry animal husbandry fishery andwater conservancy are the foundation of the national econ-omy Coal consumption was in the second place before 1990but in 1993 2000 2005 and 2008 there was a great fall(Figure 5) in contrast coal consumption of wholesale andretail trade hotels and catering services and other sectorshad an obvious increase in these three years Considerablecontact with the change of the prices is guessed

6 Results and Discussion

Table 5 shows prediction results of BVAR models under thethree different prior distributions and the ETS models Theunit is 10000 tons in ETS models which is 10000 tons ofstandard coal in BVAR models So the results of ETS modelare multiplied by the converted coefficient of 07143 as tons ofstandard coal

The result of BVAR models under the normal-Wishartprior distribution (prior = 0) is the same as under the normal-flat prior distribution (prior = 1) Using the ETS model to

Table 5 Total coal consumption forecast

BVARPrior = 0 or 1

BVARPrior = 2

ETSaggregation

ETSforecast

2011 23803337 23803337 234876 2344752012 25268675 25403407 247099 2459212013 26815268 25766741 259324 2573662014 28449414 25711339 271551 2688112015 3017629 27462729 283782 2802562016 32001513 31472065 296014 2917012017 33931196 36070012 308250 3031462018 35972003 40119003 320489 3145912019 38131212 44491339 332729 3260372020 40416784 50482369 344972 337481Unit 10000 tons of standard coal

predict coal consumption of every sector and then addingthe seven values to get 344972 billion tons of standard coalin 2020 only based on the historical total coal consumption337481 billion tons of standard coal are obtained in 2020directly FromTable 5 in 2015 forecast results of fourmethodshave small differences But in 2020 the results of BVARmodelare larger than ETS model (Figure 6)

The reasonmay be that in BVARmodel the gross value ofindustrial output and the downstream industrial productionare influencing factors whose values are increasing over the

Discrete Dynamics in Nature and Society 9

20

25

30

35

40

45

50

55

2011 2012 2013 2014 2015 2016 2017 2018 2019 2020

BVAR 0 1BVAR 2

ETS AETS F

Coa

l con

sum

ptio

n (te

n th

ousa

nd to

ns o

f sta

ndar

d co

al)

times104

Figure 6 Total coal consumption forecast

years So they drive the rise in total coal consumption in thelong run However in ETS model total coal consumption isthe sum of coal consumption in seven sectors and the abovehas introduced that the coal consumption of some sectorshas showed a decreasing trend such as agriculture forestryanimal husbandry fishery and water conservancy transportstorage and post So the ETS model prediction results aresmaller

7 Conclusions

China is a big developing country in the process of industri-alization and urbanization and therefore coal consumptionwill remain high in the future for a long time Based on guar-anteeing the stability of the VAR system coal consumptiongross industrial output value and the downstream industrialproduction (crude steel cement thermal power) are selectedas variables to establish the BVAR (2) model At the sametime this paper analyses the changes of coal consumptionstructure based on sector division The forecast models areselected through comparing information criteria and forecastaccuracy comparison from thirty ETSmodels By forecastingthe total coal consumption and analyzing structural changesthis paper gets the following main conclusions

(1) Results of the impulse response function indicate thatthe increased output of the downstream industry willbecome a driving force of coal consumption In thefirst two periods the same response has been therewith the increase of the gross value of industrialoutput Later there will be a negative response

(2) The variance decomposition results show that thecontribution of gross value of industrial outputaccounted for 50of growth in coal consumption andhas long-term stable influence in the downstreamindustry the contribution rate of crude steel outputis greater than that of the cement output and thermalpower output

(3) In the next decade from a sector division point ofview industry and household consumption account

for the largest proportion of total coal consumptionthe proportion of industrial coal consumption willcontinue to rise but the proportion of householdcoal consumption will continue to decline Transportstorage and post sectors were the largest consumersof coal consumption in 1980 which will becomethe lowest in the future Construction sector despiteits lower reliance on coal has experienced a risein coal consumption in recent years The scales ofconsumption of wholesale and retail trade hotelsand catering services agriculture forestry animalhusbandry fishery and water conservancy sectorswere influenced by price fluctuations in the past fewdecades causing the fluctuation of coal consumptionBut in the next decade coal consumption of bothsectors will have a rising trend

(4) TheBVAR (2) forecastmodels predict that in 2015 coaldemand will reach 275ndash302 billion tons of standardcoal and in 2020 it will reach 404ndash505 billion tonsof standard coal in China That is to say as themomentumof rapid development of Chinarsquos economyin the future the coal consumption in 2020 is almostdouble that of in 2012 which is 246ndash253 billion tonsof standard coal However because the unrenewablecharacteristics of the coal and the pollution causedby burning are bound to draw the attention of thecountry and the residents how to improve the coalutilization efficiency and promote the use of cleanenergy will become a main research direction in thefuture

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This research is supported by the National Natural ScienceFoundation of China (no 71103115) the China Postdoc-toral Science Foundation surface of funded projects (no2012M510580) the Fundamental Research Funds for theShaanxi province soft science (no 2012KRM95) The authorswould like to thank the anonymous referees

References

[1] L Suganthi and A A Samuel ldquoEnergy models for demandforecastingmdasha reviewrdquo Renewable and Sustainable EnergyReviews vol 16 no 2 pp 1223ndash1240 2012

[2] V S Ediger and S Akar ldquoARIMA forecasting of primary energydemand by fuel in Turkeyrdquo Energy Policy vol 35 no 3 pp 1701ndash1708 2007

[3] S HMohr and GM Evans ldquoForecasting coal production until2100rdquo Fuel vol 88 no 11 pp 2059ndash2067 2009

[4] A Unler ldquoImprovement of energy demand forecasts usingswarm intelligence the case of Turkey with projections to 2025rdquoEnergy Policy vol 36 no 6 pp 1937ndash1944 2008

10 Discrete Dynamics in Nature and Society

[5] C-C Lee and Y-B Chiu ldquoModeling OECD energy demand aninternational panel smooth transition error-correction modelrdquoInternational Review of Economics amp Finance vol 25 pp 372ndash383 2013

[6] H Chan and S Lee ldquoForecasting the demand for energy inChinardquo Energy Journal vol 17 no 1 pp 19ndash30 1996

[7] C V Hirschhausen and M Andres ldquoLong-term electricitydemand in Chinamdashfrom quantitative to qualitative growthrdquoEnergy Policy vol 28 no 4 pp 231ndash241 2000

[8] S-W Yu and K-J Zhu ldquoA hybrid procedure for energy demandforecasting in Chinardquo Energy vol 37 no 1 pp 396ndash404 2012

[9] S W Yu Y-M Wei and K Wang ldquoChinas primary energydemands in 2020 predictions from an MPSO-RBF estimationmodelrdquo Energy Conversion andManagement vol 61 pp 59ndash662012

[10] M Zhang H L Mu G Li and Y D Ning ldquoForecasting thetransport energy demand based on PLSR method in ChinardquoEnergy vol 34 no 9 pp 1396ndash1400 2009

[11] P Crompton and Y RWu ldquoEnergy consumption in China pasttrends and future directionsrdquo Energy Economics vol 27 no 1pp 195ndash208 2005

[12] J Chai Z Y Zhang J L Fu J E Guo and S Y WangldquoIdentification and analysis of the structural mutations in oilprice systemrdquo Journal of Management Sciences vol 27 no 2 pp133ndash144 2014 (Chinese)

[13] BM Francis L Moseley and S O Iyare ldquoEnergy consumptionand projected growth in selected Caribbean countriesrdquo EnergyEconomics vol 29 no 6 pp 1224ndash1232 2007

[14] L M Saini ldquoPeak load forecasting using Bayesian regulariza-tion Resilient and adaptive backpropagation learning basedartificial neural networksrdquo Electric Power Systems Research vol78 no 7 pp 1302ndash1310 2008

[15] P Lauret E Fock R N Randrianarivony and J-F Manicom-Ramsamy ldquoBayesian neural network approach to short timeload forecastingrdquo Energy Conversion and Management vol 49no 5 pp 1156ndash1166 2008

[16] C A Sims ldquoMacroeconomics and realityrdquo Econometrica vol48 no 1 pp 1ndash48 1980

[17] T Zha ldquoA dynamic multivariate model for use in formulatingpolicyrdquo Economic Review vol 83 no 1 pp 16ndash29 1998

[18] R B Litterman ldquoA Bayesian procedure for forecasting withvector autoregressionsrdquo Department of Economics WorkingPaper Massachusetts Institute of Technology 1980

[19] R B Litterman ldquoForecasting with Bayesian vector autoregres-sions five years of experiencerdquo Journal of Business and EconomicStatistics vol 4 no 1 pp 25ndash38 1986

[20] T Doan R B Litterman and C A Sims ldquoForecastingand conditional projection using realistic prior distributionsrdquoEconometric Reviews vol 3 no 1 pp 1ndash100 1984

[21] C A Sims ldquoA nine-variable probabilistic macroeconomic fore-casting modelrdquo in Business Cycles Indicators and Forecasting JH Stock and M W Watson Eds University of Chicago PressChicago Ill USA 1992

[22] C A Sims and T A Zha ldquoBayesian methods for dynamicmultivariate modelsrdquo International Economic Review vol 39no 4 pp 949ndash968 1998

[23] R J Hyndman and Y Khandakar ldquoAutomatic time seriesforecasting the forecast package for Rrdquo Journal of StatisticalSoftware vol 27 no 3 pp 1ndash22 2008

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical Problems in Engineering

Hindawi Publishing Corporationhttpwwwhindawicom

Differential EquationsInternational Journal of

Volume 2014

Applied MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical PhysicsAdvances in

Complex AnalysisJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OptimizationJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Operations ResearchAdvances in

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Function Spaces

Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of Mathematics and Mathematical Sciences

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Algebra

Discrete Dynamics in Nature and Society

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Decision SciencesAdvances in

Discrete MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Stochastic AnalysisInternational Journal of

4 Discrete Dynamics in Nature and Society

Table 1 Classification and summary of ETS models

Trend component Seasonal componentN (none) A (additive) M (multiplicative)

N (none) N N N A N MA (additive) A N A A A MAd (additive damped) Ad N Ad A Ad MM (multiplicative) M N M A M MMd (multiplicativedamped) Md N Md A Md M

error forms the above fifteen models can be extended tothirty types From the forecast results alone addition formor multiplication form hardly influence residual error termHowever for different sampled data various residual errorforms seem superior or inferior Because dividing by 0 isinvolved attention should be paid to select different additionor multiplication forms for residual error trend term orseason term When samples are all positive values usingresidual error multiplication form presents an advantageHowever when samples are zero value or negative valuemultiplication form model is not applicable

4 Forecast of Total Coal Consumption byBVAR Models

In order to analyze the dynamic relationship between coalconsumption and economic variables 5 d VARmodel is builtbased on (1) According to the lag length criteria (likelihoodratio final prediction error AIC SC and Hannan-Quinninformation criterion) VAR (2) is the most appropriatemodel VAR models are set up as follows

[[[[[

[

LNCOALLNINDLNCEMLNSTELNPOW

]]]]]

]

=

[[[[[

[

44768

minus200313

31506

minus24537

23387

]]]]]

]

+

[[[[[

[

06785 01126 minus01174 00534 04913

01305 05586 14616 minus04848 06649

minus06834 00470 10756 00198 06238

04434 minus02351 00606 08957 00600

minus02571 00139 01578 03482 09504

]]]]]

]

times

[[[[[

[

LNCOAL (minus1)LNIND (minus1)

LNCEM (minus1)

LNSTE (minus1)LNPOW (minus1)

]]]]]

]

00

05

10

15

00 05 10 15

minus05

minus10

minus15

minus05minus10minus15

Figure 1 Inverse roots of AR characteristic polynomial

+

[[[[[

[

minus03453 minus01248 minus00067 01699 minus02082

20973 minus04018 minus03793 minus05128 minus04853

03309 minus00134 minus07911 minus00454 03111

minus02974 00022 minus00554 minus02683 07020

minus00600 minus00030 minus01852 minus01937 00594

]]]]]

]

times

[[[[[

[

LNCOAL (minus2)LNIND (minus2)

LNCEM (minus2)

LNSTE (minus2)LNPOW (minus2)

]]]]]

]

+

[[[[[

[

1205761

1205762

1205763

1205764

1205765

]]]]]

]

(2)

The R-squared of the models are 09981 09951 09966and 09986 which indicate good simulation results At thesame time the reciprocals of all roots locate inside the unitcircle when calculating the AR characteristic polynomial(Figure 1) which shows that the established VAR (2) modelis stable That is to say when a variable changes in the model(ie to generate a shock) it leads to changes in other variablestoo But as time goes on the effect gradually disappears InFigure 1 the horizontal axis represents the real number axisthe vertical axis represents the imaginary numbers axis theunit circle is a circle of radius 1 and center at the origin thepoints in the unit circle represent roots of the characteristicequation on VAR (2) model

Based on the VAR model the dynamic relationshipbetween the variables can be analyzed by impulse responsefunction and variance decomposition

In Figure 2 the horizontal axis represents the lag periodsof shock from innovation (unit years) the vertical axisrepresents the response degree from the dependent variablesto the explaining variables the solid line is the calculatedvalue of the impulse response function The lag period is setto 15 years

Discrete Dynamics in Nature and Society 5

000

001

002

003

004

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

minus001

minus003

minus002

minus004

LNCOALLNINDLNCEM

LNSTELNPOW

Figure 2 Responses of LNCOAL to LNCOAL LNIND LNCEMLNSTE and LNPOW

According to Figure 2 the line graph shows that when theexplaining variables (coal consumption gross value of indus-trial output cement output crude steel industrial output andthermal power output) are shocked from innovation respec-tively how does the dependent variable (coal consumption)respond during the future 15 periods That is to say whenthese five factors change respectively we can forecast howmuch and how long will coal consumption respond

Firstly coal consumption volatility responds strongly toits own shock which is shown by the blue line and inthe first 15 periods it shows a fluctuating trend Early asa strong positive response in the second period it reachesthe peak at 00310 and then plunges to the lowest point atminus00145 in the eighth period From this point onwards theperiod between 9 and 15 experiences a gradual growth untilthe numbers become positive response again This showsthat coal consumption and its lagging values have a strongcorrelation Secondly after suffering a positive shock grossvalue of industrial output only brings positive response tocoal consumption in the second period and then it quicklydrops to a negative response to the lowest point of minus00379in the sixth period Afterwards it rises to zero slightly Thisindicates that the increase of the gross value of industrialoutput immediately leads to an increase in coal consumptionbut in the long term it reduces consumption The reasonsmay be that more output provides more capital for inter-nal restructuring and technological progress thus reducingcoal consumption costs and increasing profits Thirdly coalconsumption volatility responds modestly to cement outputshocks in the early stage From the fifth period to theeleventh period the response shows a large fluctuation fromnegative to positive Fourthly coal consumption volatilityhas increasing positive response to crude steel industrialoutput shocks gradually over time and in the seventh periodthat reaches the peak of 00219 Fifthly a positive shockon thermal power output brings positive response to coalconsumption which has weak intensity and stable tendencyThese results indicate that the development of downstream

0

20

40

60

80

100

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

LNCOALLNINDLNCEM

LNSTELNPOW

Figure 3 Factors variance decomposition on the effect of LNCOAL

industries drives an increase in demand for coal whose long-term relation is in the same direction

In Figure 3 the horizontal axis represents the lag periodsof impact from innovation (unit years) the vertical axisrepresents the contribution rate to coal consumption from theeconomic variables

As shown in Figure 3 in the future 15 periods coalconsumption has the greatest impact during the first fiveperiods with its own contribution rate above 40 Since thenrates in the rest of the period level off at about 22 Thecontribution rate of the gross value of industrial output to coalconsumption increases steeply from the third period and thenexceeds all other variables at the sixth period and remainsstable and high between 51 and 53Thermal power industryhad the highest contribution in the second period but afterthat its contribution rates decline to 3-4 Contribution ratesof crude steel industry have obvious rise from the thirdperiod and make the largest contribution in the downstreamindustries between 12ndash14 The contribution rate of cementindustry has an obvious rise in the sixth period after thatthe contribution rate remains in the 7-8 level Therefore inthe long run industrial output has the biggest contributionto the changes of coal consumption in a short time coalconsumption have the biggest contribution to own changes

The system of priors commonly used in the specificationof BVAR models includes conjugate prior distribution max-imum entropy prior distribution ML-II prior distributionand multilayer prior distribution Different BVAR priordistributions have different impacts on the prediction resultsThis paper uses the 119877 software to forecast coal consumptionbased on three kinds of prior distribution in the MSBVARpackage Command is run as follows Rgt fcastlt -szbvar (coal119901 = 2 lambda 0 = 06 lambda 1 = 01 lambda 3 = 2 lambda4 = 025 lambda 5 = 0 mu 5 = 0 mu 6 = 0 prior = 0)Thereare three different prior distributions 0 = Normal-Wishartprior 1 = Normal-flat prior and 2 = flat-flat prior (ie akintoMLE)The forecasting results from 2012 to 2020 are shownin Table 4

6 Discrete Dynamics in Nature and Society

6000

8000

10000

12000

14000

16000

18000

050000

100000150000200000250000300000350000400000450000500000

1980

1985

1990

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

2018

2019

2020

Industry coal consumptionHousehold coal consumption

Coa

l con

sum

ptio

n (1

0000

tons

)

Coa

l con

sum

ptio

n (1

0000

tons

)

Figure 4 Changes of industry and household coal consumption

5 Structural Analysis of TotalCoal Consumption

A univariate ETS (exponential smoothing) forecast model isused to predict coal consumption of each sector in 2020 inthe forecast package for R Error Trend and Seasonal of ETSmodel denote error term (A M Z) trend term (N A M Z)and season term (NAM Z) respectively inwhich the groupincludes thirty models

From the coal balance sheet in China Statistical YearbookandWINDDatabase coal consumption data of seven sectorsare selected for the period 1980ndash2010 The balance equa-tion for coal consumption Total coal consumption (TO) =Agriculture Forestry Animal Husbandry Fishery andWaterConservancy (AFAFW) + Industry (IND) + Construction(CON) + Transport Storage and Post (TSP) +Wholesale andRetail Trades Hotels and Catering Services (WRHC) +OtherSectors (OS) + Household Consumption (HC) Unit is 10000tons

The finalmodels (Table 2) are selected from thirtymodelsby comparing information criterion (AIC BIC AICc) andforecast precision (mean error root mean square error meanabsolute error maximum permissible errors mean absolutepercentage error mean absolute error square) (Table 3)Command is run as follows R gt fcast lt -forecast(est(x))

As can be seen from the Table 4 the total coal consump-tion is the sum of coal consumption in seven sectors whichis 39728663 billion tons in 2015 This value is similar tothe goal in the development of coal industry in the twelfthfive-year plan which is 39 billion tons in 2015 By 2020Chinarsquos total coal consumption will reach 483 billion tonswith an annual growth rate of 436 of which industry andhousehold consumption account for the largest proportionIn Figure 4 the industry coal consumption was 438 milliontons 811 million tons 1278 billion tons and 296 billiontons in 1980 1990 2000 and 2010 respectively and it ispredicted to reach 4669 billion tons in 2020 accountingfor 9667 of total coal consumption Therefore to reducecoal consumption in China industrial energy conservationand emissions reduction are the fastest and most effective

breakthrough points including the adjustment of industrialstructure and optimization of technology

However as the second largest main channel the propor-tion of household consumption is decreasing year by yearFigure 4 shows that household consumption was up to 167million tons in 1990 which accounts for 1583 of the totalcoal consumption By 2020 household consumption willbe 092 million tons accounting for 189 of the total coalconsumption In the field of household energy consumptionas the diversity of the type of energy and development oftechnology electricity and heating oil are more available andconvenient than coal for residents so less and less coal is usedin the household rapidly This is beneficial to environmentalprotection and nonrenewable resources saving

As shown in Figure 5 transport storage and post sectorswere the largest coal consumption sectors in 1980 but theybecame the lowest coal consumption sectors after thirty yearsin 2010 and their coal consumption will decline constantlyduring the next decade by predictionsThemain reason is thatcoal-fired steam locomotive engines have been replaced byinternal combustion and electric locomotives by the railwaysSo most of the raw coal has been replaced by fuel andelectricity at the same time Nowadays road transporta-tion mainly consumes gasoline and diesel oil waterwaystransportation mainly consumes fuel oil and diesel oil airtransport mainly consumes aviation kerosene Thereforealthough the development of transport storage and postsectors still consumes large amounts of energy the usage ofcoal is insignificant

As the pillar industry of national economy in our countryconstruction sector is also an energy-intensive industryCoal consumption increased from 556 to 71892 milliontons between 1980 and 2010 representing a relatively smallgrowth By forecasting the amount of coal consumption isstill increasing in the future (Figure 5) At the same time theconstruction sector does not have a high proportion of coalconsumption and accounted for only 023 in 2010 That isbecause there are various types of energies consumed by theconstruction industry including raw coal gasoline diesel oilfuel oil heat electricity and other petroleum products With

Discrete Dynamics in Nature and Society 7

300

600

900

1200

1500

1800

2100

2400

2700

1980

1985

1990

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

2018

2019

2020

AFAFWCONTSP

WRHCOS

Coa

l con

sum

ptio

n (1

0000

tons

)

Figure 5 Changes of other sectorsrsquo coal consumption

Table 2 Formulae for recursive calculations point forecasts and parametric estimated value on sector by ETS method

Variables Forms Models and parameters

AFAFW (M N N)119897119905= 120572119910119905+ (1 minus 120572) 119897

119905minus1 119910119905+ℎ|119905

= 119897119905

119910119905+1

= 119910119905+1|119905

(1 + 120576119905) 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 00001 119897 = 17937859 120590 = 01809

IND (A A N)

119897119905= 120572119910119905+ (1 minus 120572) (119897

119905minus1+ 119887119905minus1) 119887119905= 120573 (119897

119905minus 119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905+ ℎ119887119905

119910119905+1

= 119910119905+1|119905

+ 120576119905 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 09999 120573 = 02697119897 = 547821109 119887 = 93385498 120590 = 129103600

CON (A A N)119897119905= 120572119910119905+ (1 minus 120572) (119897

119905minus1+ 119887119905minus1) 119887119905= 120573 (119897

119905minus 119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905+ ℎ119887119905

119910119905+1

= 119910119905+1|119905

+ 120576119905 120576119905sim 119873119868119863 (0 120590

2) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 00009 120573 = 00009 119897 = 4483868 119887 = 87382 120590 = 536530

TSP (M M N)119897119905= 119886119910119905+ (1 minus 119886) 119897

119905minus1119887119905minus1

119887119905= 120573 (119897

119905119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905119887119905

ℎ

119910119905+1

= 119910119905+1|119905

(1 + 120576119905) 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 00001 120573 = 00001 119897 = 23050531 119887 = 09371 120590 = 01049

WRHC (M M N)119897119905= 119886119910119905+ (1 minus 119886) 119897

119905minus1119887119905minus1

119887119905= 120573 (119897

119905119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905119887119905

ℎ

119910119905+1

= 119910119905+1|119905

(1 + 120576119905) 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 00001 120573 = 00001 119897 = 7651436 119887 = 10328 120590 = 02562

OS (A N N)119897119905= 119886119910119905+ (1 minus 119886) 119897

119905minus1 119910119905+ℎ|119905

= 119897119905

119910119905+1

= 119910119905+1|119905

+ 120576119905 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 08917 119897 = 11487978 120590 = 4978256

HC (M N N)119897119905= 119886119910119905+ (1 minus 119886) 119897

119905minus1 119910119905+ℎ|119905

= 119897119905

119910119905+1

= 119910119905+1|119905

(1 + 120576119905) 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 09999 119897 = 113736307 120590 = 01327

TO (A A N)119897119905= 120572119910119905+ (1 minus 120572) (119897

119905minus1+ 119887119905minus1) 119887119905= 120573 (119897

119905minus 119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905+ ℎ119887119905

119910119905+1

= 119910119905+1|119905

+ 120576119905 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 09999 120573 = 01665 119897 = 784219911 119887 = 93651674 120590 = 146462100

In the above formulae for recursive calculations and point forecasts 119897119905denotes the series level at time 119905 119887

119905denotes the

slope at time 119905 120572 and 120573 are smoothness index ℎ denotes the lag period

the development of technology the construction sector willbecome dependent on power and all kinds of oil productsincreasingly instead of coal which can improve the energyefficiency and control environmental pollution gradually

Wholesale and retail trade and hotels and catering ser-vices consume coal electricity heat gasoline diesel oil lique-fied petroleum gas natural gas and other varieties of energy

Influenced by price changes and economic development thissector has gone through booms and busts repeatedly From1980 to 1990 the sector had a rising trend due to adaptingto the market demand quickly from 1990 to 1997 with highprice the sector was in the integration stage from 1997 to1999 the sector was in a low price and low growth stagefrom 2000 to 2004 the sector was in a low price and high

8 Discrete Dynamics in Nature and Society

Table 3 Accuracy of forecast models on sector by ETS method

AFAFW IND CON TSP WRHC OS HC TOAIC 3107869 4694980 2392015 2724609 3085274 3287655 3715486 4747963AICc 3114536 4719980 2417015 2749609 3110274 3294321 3722153 4772963BIC 3128760 4736761 2433796 2766390 3127055 3308545 3736377 4789744ME minus01636 12971479 02783 53547 minus218557 455007 minus1054612 19036901RMSE 3244691 129103640 536530 1544569 3098943 4978256 15404774 146462092MAE 2556755 106357804 396210 1158266 2727594 3650402 10815817 118112234MPE minus40490 minus02898 minus10556 minus08693 minus105543 minus64377 minus19142 minus01609MAPE 157232 98982 77865 87576 271766 300246 96963 93149MASE 07907 07024 08702 07315 11268 09915 09609 07435

Table 4 Forecasting result of coal consumption on sector by ETS method

ETS TO AFAFW IND CON TSP WRHC OS HC2011 32882036 179379 31311760 64068 55260 155568 200086 9159172012 34593125 179379 33020350 64942 51787 160666 200086 9159172013 36304590 179379 34728930 65816 48532 165931 200086 9159172014 38016441 179379 36437520 66691 45481 171368 200086 9159172015 39728663 179379 38146110 67565 42623 176984 200086 9159172016 41441238 179379 39854690 68439 39944 182784 200086 9159172017 43154182 179379 41563280 69314 37433 188774 200086 9159172018 44867480 179379 43271870 70188 35080 194961 200086 9159172019 46581118 179379 44980450 71062 32875 201350 200086 9159172020 48295114 179379 46689040 71937 30809 207948 200086 915917Unit 10000 tons

growth stage domestic prices began to soar in 2005 andsince the American subprime crisis and the domestic policyadjustment in 2008 the sector appears to have had violentfluctuations However Figure 5 shows that the wholesale andretail trade hotels and catering services will have a growthtrend of coal consumption in the future Industry scale willcontinue to expand which is consistent with our countryrsquospolicy to develop the tertiary industry vigorously

Agriculture forestry animal husbandry fishery andwater conservancy are the foundation of the national econ-omy Coal consumption was in the second place before 1990but in 1993 2000 2005 and 2008 there was a great fall(Figure 5) in contrast coal consumption of wholesale andretail trade hotels and catering services and other sectorshad an obvious increase in these three years Considerablecontact with the change of the prices is guessed

6 Results and Discussion

Table 5 shows prediction results of BVAR models under thethree different prior distributions and the ETS models Theunit is 10000 tons in ETS models which is 10000 tons ofstandard coal in BVAR models So the results of ETS modelare multiplied by the converted coefficient of 07143 as tons ofstandard coal

The result of BVAR models under the normal-Wishartprior distribution (prior = 0) is the same as under the normal-flat prior distribution (prior = 1) Using the ETS model to

Table 5 Total coal consumption forecast

BVARPrior = 0 or 1

BVARPrior = 2

ETSaggregation

ETSforecast

2011 23803337 23803337 234876 2344752012 25268675 25403407 247099 2459212013 26815268 25766741 259324 2573662014 28449414 25711339 271551 2688112015 3017629 27462729 283782 2802562016 32001513 31472065 296014 2917012017 33931196 36070012 308250 3031462018 35972003 40119003 320489 3145912019 38131212 44491339 332729 3260372020 40416784 50482369 344972 337481Unit 10000 tons of standard coal

predict coal consumption of every sector and then addingthe seven values to get 344972 billion tons of standard coalin 2020 only based on the historical total coal consumption337481 billion tons of standard coal are obtained in 2020directly FromTable 5 in 2015 forecast results of fourmethodshave small differences But in 2020 the results of BVARmodelare larger than ETS model (Figure 6)

The reasonmay be that in BVARmodel the gross value ofindustrial output and the downstream industrial productionare influencing factors whose values are increasing over the

Discrete Dynamics in Nature and Society 9

20

25

30

35

40

45

50

55

2011 2012 2013 2014 2015 2016 2017 2018 2019 2020

BVAR 0 1BVAR 2

ETS AETS F

Coa

l con

sum

ptio

n (te

n th

ousa

nd to

ns o

f sta

ndar

d co

al)

times104

Figure 6 Total coal consumption forecast

years So they drive the rise in total coal consumption in thelong run However in ETS model total coal consumption isthe sum of coal consumption in seven sectors and the abovehas introduced that the coal consumption of some sectorshas showed a decreasing trend such as agriculture forestryanimal husbandry fishery and water conservancy transportstorage and post So the ETS model prediction results aresmaller

7 Conclusions

China is a big developing country in the process of industri-alization and urbanization and therefore coal consumptionwill remain high in the future for a long time Based on guar-anteeing the stability of the VAR system coal consumptiongross industrial output value and the downstream industrialproduction (crude steel cement thermal power) are selectedas variables to establish the BVAR (2) model At the sametime this paper analyses the changes of coal consumptionstructure based on sector division The forecast models areselected through comparing information criteria and forecastaccuracy comparison from thirty ETSmodels By forecastingthe total coal consumption and analyzing structural changesthis paper gets the following main conclusions

(1) Results of the impulse response function indicate thatthe increased output of the downstream industry willbecome a driving force of coal consumption In thefirst two periods the same response has been therewith the increase of the gross value of industrialoutput Later there will be a negative response

(2) The variance decomposition results show that thecontribution of gross value of industrial outputaccounted for 50of growth in coal consumption andhas long-term stable influence in the downstreamindustry the contribution rate of crude steel outputis greater than that of the cement output and thermalpower output

(3) In the next decade from a sector division point ofview industry and household consumption account

for the largest proportion of total coal consumptionthe proportion of industrial coal consumption willcontinue to rise but the proportion of householdcoal consumption will continue to decline Transportstorage and post sectors were the largest consumersof coal consumption in 1980 which will becomethe lowest in the future Construction sector despiteits lower reliance on coal has experienced a risein coal consumption in recent years The scales ofconsumption of wholesale and retail trade hotelsand catering services agriculture forestry animalhusbandry fishery and water conservancy sectorswere influenced by price fluctuations in the past fewdecades causing the fluctuation of coal consumptionBut in the next decade coal consumption of bothsectors will have a rising trend

(4) TheBVAR (2) forecastmodels predict that in 2015 coaldemand will reach 275ndash302 billion tons of standardcoal and in 2020 it will reach 404ndash505 billion tonsof standard coal in China That is to say as themomentumof rapid development of Chinarsquos economyin the future the coal consumption in 2020 is almostdouble that of in 2012 which is 246ndash253 billion tonsof standard coal However because the unrenewablecharacteristics of the coal and the pollution causedby burning are bound to draw the attention of thecountry and the residents how to improve the coalutilization efficiency and promote the use of cleanenergy will become a main research direction in thefuture

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This research is supported by the National Natural ScienceFoundation of China (no 71103115) the China Postdoc-toral Science Foundation surface of funded projects (no2012M510580) the Fundamental Research Funds for theShaanxi province soft science (no 2012KRM95) The authorswould like to thank the anonymous referees

References

[1] L Suganthi and A A Samuel ldquoEnergy models for demandforecastingmdasha reviewrdquo Renewable and Sustainable EnergyReviews vol 16 no 2 pp 1223ndash1240 2012

[2] V S Ediger and S Akar ldquoARIMA forecasting of primary energydemand by fuel in Turkeyrdquo Energy Policy vol 35 no 3 pp 1701ndash1708 2007

[3] S HMohr and GM Evans ldquoForecasting coal production until2100rdquo Fuel vol 88 no 11 pp 2059ndash2067 2009

[4] A Unler ldquoImprovement of energy demand forecasts usingswarm intelligence the case of Turkey with projections to 2025rdquoEnergy Policy vol 36 no 6 pp 1937ndash1944 2008

10 Discrete Dynamics in Nature and Society

[5] C-C Lee and Y-B Chiu ldquoModeling OECD energy demand aninternational panel smooth transition error-correction modelrdquoInternational Review of Economics amp Finance vol 25 pp 372ndash383 2013

[6] H Chan and S Lee ldquoForecasting the demand for energy inChinardquo Energy Journal vol 17 no 1 pp 19ndash30 1996

[7] C V Hirschhausen and M Andres ldquoLong-term electricitydemand in Chinamdashfrom quantitative to qualitative growthrdquoEnergy Policy vol 28 no 4 pp 231ndash241 2000

[8] S-W Yu and K-J Zhu ldquoA hybrid procedure for energy demandforecasting in Chinardquo Energy vol 37 no 1 pp 396ndash404 2012

[9] S W Yu Y-M Wei and K Wang ldquoChinas primary energydemands in 2020 predictions from an MPSO-RBF estimationmodelrdquo Energy Conversion andManagement vol 61 pp 59ndash662012

[10] M Zhang H L Mu G Li and Y D Ning ldquoForecasting thetransport energy demand based on PLSR method in ChinardquoEnergy vol 34 no 9 pp 1396ndash1400 2009

[11] P Crompton and Y RWu ldquoEnergy consumption in China pasttrends and future directionsrdquo Energy Economics vol 27 no 1pp 195ndash208 2005

[12] J Chai Z Y Zhang J L Fu J E Guo and S Y WangldquoIdentification and analysis of the structural mutations in oilprice systemrdquo Journal of Management Sciences vol 27 no 2 pp133ndash144 2014 (Chinese)

[13] BM Francis L Moseley and S O Iyare ldquoEnergy consumptionand projected growth in selected Caribbean countriesrdquo EnergyEconomics vol 29 no 6 pp 1224ndash1232 2007

[14] L M Saini ldquoPeak load forecasting using Bayesian regulariza-tion Resilient and adaptive backpropagation learning basedartificial neural networksrdquo Electric Power Systems Research vol78 no 7 pp 1302ndash1310 2008

[15] P Lauret E Fock R N Randrianarivony and J-F Manicom-Ramsamy ldquoBayesian neural network approach to short timeload forecastingrdquo Energy Conversion and Management vol 49no 5 pp 1156ndash1166 2008

[16] C A Sims ldquoMacroeconomics and realityrdquo Econometrica vol48 no 1 pp 1ndash48 1980

[17] T Zha ldquoA dynamic multivariate model for use in formulatingpolicyrdquo Economic Review vol 83 no 1 pp 16ndash29 1998

[18] R B Litterman ldquoA Bayesian procedure for forecasting withvector autoregressionsrdquo Department of Economics WorkingPaper Massachusetts Institute of Technology 1980

[19] R B Litterman ldquoForecasting with Bayesian vector autoregres-sions five years of experiencerdquo Journal of Business and EconomicStatistics vol 4 no 1 pp 25ndash38 1986

[20] T Doan R B Litterman and C A Sims ldquoForecastingand conditional projection using realistic prior distributionsrdquoEconometric Reviews vol 3 no 1 pp 1ndash100 1984

[21] C A Sims ldquoA nine-variable probabilistic macroeconomic fore-casting modelrdquo in Business Cycles Indicators and Forecasting JH Stock and M W Watson Eds University of Chicago PressChicago Ill USA 1992

[22] C A Sims and T A Zha ldquoBayesian methods for dynamicmultivariate modelsrdquo International Economic Review vol 39no 4 pp 949ndash968 1998

[23] R J Hyndman and Y Khandakar ldquoAutomatic time seriesforecasting the forecast package for Rrdquo Journal of StatisticalSoftware vol 27 no 3 pp 1ndash22 2008

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical Problems in Engineering

Hindawi Publishing Corporationhttpwwwhindawicom

Differential EquationsInternational Journal of

Volume 2014

Applied MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical PhysicsAdvances in

Complex AnalysisJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OptimizationJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Operations ResearchAdvances in

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Function Spaces

Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of Mathematics and Mathematical Sciences

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Algebra

Discrete Dynamics in Nature and Society

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Decision SciencesAdvances in

Discrete MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Stochastic AnalysisInternational Journal of

Discrete Dynamics in Nature and Society 5

000

001

002

003

004

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

minus001

minus003

minus002

minus004

LNCOALLNINDLNCEM

LNSTELNPOW

Figure 2 Responses of LNCOAL to LNCOAL LNIND LNCEMLNSTE and LNPOW

According to Figure 2 the line graph shows that when theexplaining variables (coal consumption gross value of indus-trial output cement output crude steel industrial output andthermal power output) are shocked from innovation respec-tively how does the dependent variable (coal consumption)respond during the future 15 periods That is to say whenthese five factors change respectively we can forecast howmuch and how long will coal consumption respond

Firstly coal consumption volatility responds strongly toits own shock which is shown by the blue line and inthe first 15 periods it shows a fluctuating trend Early asa strong positive response in the second period it reachesthe peak at 00310 and then plunges to the lowest point atminus00145 in the eighth period From this point onwards theperiod between 9 and 15 experiences a gradual growth untilthe numbers become positive response again This showsthat coal consumption and its lagging values have a strongcorrelation Secondly after suffering a positive shock grossvalue of industrial output only brings positive response tocoal consumption in the second period and then it quicklydrops to a negative response to the lowest point of minus00379in the sixth period Afterwards it rises to zero slightly Thisindicates that the increase of the gross value of industrialoutput immediately leads to an increase in coal consumptionbut in the long term it reduces consumption The reasonsmay be that more output provides more capital for inter-nal restructuring and technological progress thus reducingcoal consumption costs and increasing profits Thirdly coalconsumption volatility responds modestly to cement outputshocks in the early stage From the fifth period to theeleventh period the response shows a large fluctuation fromnegative to positive Fourthly coal consumption volatilityhas increasing positive response to crude steel industrialoutput shocks gradually over time and in the seventh periodthat reaches the peak of 00219 Fifthly a positive shockon thermal power output brings positive response to coalconsumption which has weak intensity and stable tendencyThese results indicate that the development of downstream

0

20

40

60

80

100

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

LNCOALLNINDLNCEM

LNSTELNPOW

Figure 3 Factors variance decomposition on the effect of LNCOAL

industries drives an increase in demand for coal whose long-term relation is in the same direction

In Figure 3 the horizontal axis represents the lag periodsof impact from innovation (unit years) the vertical axisrepresents the contribution rate to coal consumption from theeconomic variables

As shown in Figure 3 in the future 15 periods coalconsumption has the greatest impact during the first fiveperiods with its own contribution rate above 40 Since thenrates in the rest of the period level off at about 22 Thecontribution rate of the gross value of industrial output to coalconsumption increases steeply from the third period and thenexceeds all other variables at the sixth period and remainsstable and high between 51 and 53Thermal power industryhad the highest contribution in the second period but afterthat its contribution rates decline to 3-4 Contribution ratesof crude steel industry have obvious rise from the thirdperiod and make the largest contribution in the downstreamindustries between 12ndash14 The contribution rate of cementindustry has an obvious rise in the sixth period after thatthe contribution rate remains in the 7-8 level Therefore inthe long run industrial output has the biggest contributionto the changes of coal consumption in a short time coalconsumption have the biggest contribution to own changes

The system of priors commonly used in the specificationof BVAR models includes conjugate prior distribution max-imum entropy prior distribution ML-II prior distributionand multilayer prior distribution Different BVAR priordistributions have different impacts on the prediction resultsThis paper uses the 119877 software to forecast coal consumptionbased on three kinds of prior distribution in the MSBVARpackage Command is run as follows Rgt fcastlt -szbvar (coal119901 = 2 lambda 0 = 06 lambda 1 = 01 lambda 3 = 2 lambda4 = 025 lambda 5 = 0 mu 5 = 0 mu 6 = 0 prior = 0)Thereare three different prior distributions 0 = Normal-Wishartprior 1 = Normal-flat prior and 2 = flat-flat prior (ie akintoMLE)The forecasting results from 2012 to 2020 are shownin Table 4

6 Discrete Dynamics in Nature and Society

6000

8000

10000

12000

14000

16000

18000

050000

100000150000200000250000300000350000400000450000500000

1980

1985

1990

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

2018

2019

2020

Industry coal consumptionHousehold coal consumption

Coa

l con

sum

ptio

n (1

0000

tons

)

Coa

l con

sum

ptio

n (1

0000

tons

)

Figure 4 Changes of industry and household coal consumption

5 Structural Analysis of TotalCoal Consumption

A univariate ETS (exponential smoothing) forecast model isused to predict coal consumption of each sector in 2020 inthe forecast package for R Error Trend and Seasonal of ETSmodel denote error term (A M Z) trend term (N A M Z)and season term (NAM Z) respectively inwhich the groupincludes thirty models

From the coal balance sheet in China Statistical YearbookandWINDDatabase coal consumption data of seven sectorsare selected for the period 1980ndash2010 The balance equa-tion for coal consumption Total coal consumption (TO) =Agriculture Forestry Animal Husbandry Fishery andWaterConservancy (AFAFW) + Industry (IND) + Construction(CON) + Transport Storage and Post (TSP) +Wholesale andRetail Trades Hotels and Catering Services (WRHC) +OtherSectors (OS) + Household Consumption (HC) Unit is 10000tons

The finalmodels (Table 2) are selected from thirtymodelsby comparing information criterion (AIC BIC AICc) andforecast precision (mean error root mean square error meanabsolute error maximum permissible errors mean absolutepercentage error mean absolute error square) (Table 3)Command is run as follows R gt fcast lt -forecast(est(x))

As can be seen from the Table 4 the total coal consump-tion is the sum of coal consumption in seven sectors whichis 39728663 billion tons in 2015 This value is similar tothe goal in the development of coal industry in the twelfthfive-year plan which is 39 billion tons in 2015 By 2020Chinarsquos total coal consumption will reach 483 billion tonswith an annual growth rate of 436 of which industry andhousehold consumption account for the largest proportionIn Figure 4 the industry coal consumption was 438 milliontons 811 million tons 1278 billion tons and 296 billiontons in 1980 1990 2000 and 2010 respectively and it ispredicted to reach 4669 billion tons in 2020 accountingfor 9667 of total coal consumption Therefore to reducecoal consumption in China industrial energy conservationand emissions reduction are the fastest and most effective

breakthrough points including the adjustment of industrialstructure and optimization of technology

However as the second largest main channel the propor-tion of household consumption is decreasing year by yearFigure 4 shows that household consumption was up to 167million tons in 1990 which accounts for 1583 of the totalcoal consumption By 2020 household consumption willbe 092 million tons accounting for 189 of the total coalconsumption In the field of household energy consumptionas the diversity of the type of energy and development oftechnology electricity and heating oil are more available andconvenient than coal for residents so less and less coal is usedin the household rapidly This is beneficial to environmentalprotection and nonrenewable resources saving

As shown in Figure 5 transport storage and post sectorswere the largest coal consumption sectors in 1980 but theybecame the lowest coal consumption sectors after thirty yearsin 2010 and their coal consumption will decline constantlyduring the next decade by predictionsThemain reason is thatcoal-fired steam locomotive engines have been replaced byinternal combustion and electric locomotives by the railwaysSo most of the raw coal has been replaced by fuel andelectricity at the same time Nowadays road transporta-tion mainly consumes gasoline and diesel oil waterwaystransportation mainly consumes fuel oil and diesel oil airtransport mainly consumes aviation kerosene Thereforealthough the development of transport storage and postsectors still consumes large amounts of energy the usage ofcoal is insignificant

As the pillar industry of national economy in our countryconstruction sector is also an energy-intensive industryCoal consumption increased from 556 to 71892 milliontons between 1980 and 2010 representing a relatively smallgrowth By forecasting the amount of coal consumption isstill increasing in the future (Figure 5) At the same time theconstruction sector does not have a high proportion of coalconsumption and accounted for only 023 in 2010 That isbecause there are various types of energies consumed by theconstruction industry including raw coal gasoline diesel oilfuel oil heat electricity and other petroleum products With

Discrete Dynamics in Nature and Society 7

300

600

900

1200

1500

1800

2100

2400

2700

1980

1985

1990

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

2018

2019

2020

AFAFWCONTSP

WRHCOS

Coa

l con

sum

ptio

n (1

0000

tons

)

Figure 5 Changes of other sectorsrsquo coal consumption

Table 2 Formulae for recursive calculations point forecasts and parametric estimated value on sector by ETS method

Variables Forms Models and parameters

AFAFW (M N N)119897119905= 120572119910119905+ (1 minus 120572) 119897

119905minus1 119910119905+ℎ|119905

= 119897119905

119910119905+1

= 119910119905+1|119905

(1 + 120576119905) 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 00001 119897 = 17937859 120590 = 01809

IND (A A N)

119897119905= 120572119910119905+ (1 minus 120572) (119897

119905minus1+ 119887119905minus1) 119887119905= 120573 (119897

119905minus 119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905+ ℎ119887119905

119910119905+1

= 119910119905+1|119905

+ 120576119905 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 09999 120573 = 02697119897 = 547821109 119887 = 93385498 120590 = 129103600

CON (A A N)119897119905= 120572119910119905+ (1 minus 120572) (119897

119905minus1+ 119887119905minus1) 119887119905= 120573 (119897

119905minus 119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905+ ℎ119887119905

119910119905+1

= 119910119905+1|119905

+ 120576119905 120576119905sim 119873119868119863 (0 120590

2) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 00009 120573 = 00009 119897 = 4483868 119887 = 87382 120590 = 536530

TSP (M M N)119897119905= 119886119910119905+ (1 minus 119886) 119897

119905minus1119887119905minus1

119887119905= 120573 (119897

119905119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905119887119905

ℎ

119910119905+1

= 119910119905+1|119905

(1 + 120576119905) 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 00001 120573 = 00001 119897 = 23050531 119887 = 09371 120590 = 01049

WRHC (M M N)119897119905= 119886119910119905+ (1 minus 119886) 119897

119905minus1119887119905minus1

119887119905= 120573 (119897

119905119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905119887119905

ℎ

119910119905+1

= 119910119905+1|119905

(1 + 120576119905) 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 00001 120573 = 00001 119897 = 7651436 119887 = 10328 120590 = 02562

OS (A N N)119897119905= 119886119910119905+ (1 minus 119886) 119897

119905minus1 119910119905+ℎ|119905

= 119897119905

119910119905+1

= 119910119905+1|119905

+ 120576119905 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 08917 119897 = 11487978 120590 = 4978256

HC (M N N)119897119905= 119886119910119905+ (1 minus 119886) 119897

119905minus1 119910119905+ℎ|119905

= 119897119905

119910119905+1

= 119910119905+1|119905

(1 + 120576119905) 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 09999 119897 = 113736307 120590 = 01327

TO (A A N)119897119905= 120572119910119905+ (1 minus 120572) (119897

119905minus1+ 119887119905minus1) 119887119905= 120573 (119897

119905minus 119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905+ ℎ119887119905

119910119905+1

= 119910119905+1|119905

+ 120576119905 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 09999 120573 = 01665 119897 = 784219911 119887 = 93651674 120590 = 146462100

In the above formulae for recursive calculations and point forecasts 119897119905denotes the series level at time 119905 119887

119905denotes the

slope at time 119905 120572 and 120573 are smoothness index ℎ denotes the lag period

the development of technology the construction sector willbecome dependent on power and all kinds of oil productsincreasingly instead of coal which can improve the energyefficiency and control environmental pollution gradually

Wholesale and retail trade and hotels and catering ser-vices consume coal electricity heat gasoline diesel oil lique-fied petroleum gas natural gas and other varieties of energy

Influenced by price changes and economic development thissector has gone through booms and busts repeatedly From1980 to 1990 the sector had a rising trend due to adaptingto the market demand quickly from 1990 to 1997 with highprice the sector was in the integration stage from 1997 to1999 the sector was in a low price and low growth stagefrom 2000 to 2004 the sector was in a low price and high

8 Discrete Dynamics in Nature and Society

Table 3 Accuracy of forecast models on sector by ETS method

AFAFW IND CON TSP WRHC OS HC TOAIC 3107869 4694980 2392015 2724609 3085274 3287655 3715486 4747963AICc 3114536 4719980 2417015 2749609 3110274 3294321 3722153 4772963BIC 3128760 4736761 2433796 2766390 3127055 3308545 3736377 4789744ME minus01636 12971479 02783 53547 minus218557 455007 minus1054612 19036901RMSE 3244691 129103640 536530 1544569 3098943 4978256 15404774 146462092MAE 2556755 106357804 396210 1158266 2727594 3650402 10815817 118112234MPE minus40490 minus02898 minus10556 minus08693 minus105543 minus64377 minus19142 minus01609MAPE 157232 98982 77865 87576 271766 300246 96963 93149MASE 07907 07024 08702 07315 11268 09915 09609 07435

Table 4 Forecasting result of coal consumption on sector by ETS method

ETS TO AFAFW IND CON TSP WRHC OS HC2011 32882036 179379 31311760 64068 55260 155568 200086 9159172012 34593125 179379 33020350 64942 51787 160666 200086 9159172013 36304590 179379 34728930 65816 48532 165931 200086 9159172014 38016441 179379 36437520 66691 45481 171368 200086 9159172015 39728663 179379 38146110 67565 42623 176984 200086 9159172016 41441238 179379 39854690 68439 39944 182784 200086 9159172017 43154182 179379 41563280 69314 37433 188774 200086 9159172018 44867480 179379 43271870 70188 35080 194961 200086 9159172019 46581118 179379 44980450 71062 32875 201350 200086 9159172020 48295114 179379 46689040 71937 30809 207948 200086 915917Unit 10000 tons

growth stage domestic prices began to soar in 2005 andsince the American subprime crisis and the domestic policyadjustment in 2008 the sector appears to have had violentfluctuations However Figure 5 shows that the wholesale andretail trade hotels and catering services will have a growthtrend of coal consumption in the future Industry scale willcontinue to expand which is consistent with our countryrsquospolicy to develop the tertiary industry vigorously

Agriculture forestry animal husbandry fishery andwater conservancy are the foundation of the national econ-omy Coal consumption was in the second place before 1990but in 1993 2000 2005 and 2008 there was a great fall(Figure 5) in contrast coal consumption of wholesale andretail trade hotels and catering services and other sectorshad an obvious increase in these three years Considerablecontact with the change of the prices is guessed

6 Results and Discussion

Table 5 shows prediction results of BVAR models under thethree different prior distributions and the ETS models Theunit is 10000 tons in ETS models which is 10000 tons ofstandard coal in BVAR models So the results of ETS modelare multiplied by the converted coefficient of 07143 as tons ofstandard coal

The result of BVAR models under the normal-Wishartprior distribution (prior = 0) is the same as under the normal-flat prior distribution (prior = 1) Using the ETS model to

Table 5 Total coal consumption forecast

BVARPrior = 0 or 1

BVARPrior = 2

ETSaggregation

ETSforecast

2011 23803337 23803337 234876 2344752012 25268675 25403407 247099 2459212013 26815268 25766741 259324 2573662014 28449414 25711339 271551 2688112015 3017629 27462729 283782 2802562016 32001513 31472065 296014 2917012017 33931196 36070012 308250 3031462018 35972003 40119003 320489 3145912019 38131212 44491339 332729 3260372020 40416784 50482369 344972 337481Unit 10000 tons of standard coal

predict coal consumption of every sector and then addingthe seven values to get 344972 billion tons of standard coalin 2020 only based on the historical total coal consumption337481 billion tons of standard coal are obtained in 2020directly FromTable 5 in 2015 forecast results of fourmethodshave small differences But in 2020 the results of BVARmodelare larger than ETS model (Figure 6)

The reasonmay be that in BVARmodel the gross value ofindustrial output and the downstream industrial productionare influencing factors whose values are increasing over the

Discrete Dynamics in Nature and Society 9

20

25

30

35

40

45

50

55

2011 2012 2013 2014 2015 2016 2017 2018 2019 2020

BVAR 0 1BVAR 2

ETS AETS F

Coa

l con

sum

ptio

n (te

n th

ousa

nd to

ns o

f sta

ndar

d co

al)

times104

Figure 6 Total coal consumption forecast

years So they drive the rise in total coal consumption in thelong run However in ETS model total coal consumption isthe sum of coal consumption in seven sectors and the abovehas introduced that the coal consumption of some sectorshas showed a decreasing trend such as agriculture forestryanimal husbandry fishery and water conservancy transportstorage and post So the ETS model prediction results aresmaller

7 Conclusions

China is a big developing country in the process of industri-alization and urbanization and therefore coal consumptionwill remain high in the future for a long time Based on guar-anteeing the stability of the VAR system coal consumptiongross industrial output value and the downstream industrialproduction (crude steel cement thermal power) are selectedas variables to establish the BVAR (2) model At the sametime this paper analyses the changes of coal consumptionstructure based on sector division The forecast models areselected through comparing information criteria and forecastaccuracy comparison from thirty ETSmodels By forecastingthe total coal consumption and analyzing structural changesthis paper gets the following main conclusions

(1) Results of the impulse response function indicate thatthe increased output of the downstream industry willbecome a driving force of coal consumption In thefirst two periods the same response has been therewith the increase of the gross value of industrialoutput Later there will be a negative response

(2) The variance decomposition results show that thecontribution of gross value of industrial outputaccounted for 50of growth in coal consumption andhas long-term stable influence in the downstreamindustry the contribution rate of crude steel outputis greater than that of the cement output and thermalpower output

(3) In the next decade from a sector division point ofview industry and household consumption account

for the largest proportion of total coal consumptionthe proportion of industrial coal consumption willcontinue to rise but the proportion of householdcoal consumption will continue to decline Transportstorage and post sectors were the largest consumersof coal consumption in 1980 which will becomethe lowest in the future Construction sector despiteits lower reliance on coal has experienced a risein coal consumption in recent years The scales ofconsumption of wholesale and retail trade hotelsand catering services agriculture forestry animalhusbandry fishery and water conservancy sectorswere influenced by price fluctuations in the past fewdecades causing the fluctuation of coal consumptionBut in the next decade coal consumption of bothsectors will have a rising trend

(4) TheBVAR (2) forecastmodels predict that in 2015 coaldemand will reach 275ndash302 billion tons of standardcoal and in 2020 it will reach 404ndash505 billion tonsof standard coal in China That is to say as themomentumof rapid development of Chinarsquos economyin the future the coal consumption in 2020 is almostdouble that of in 2012 which is 246ndash253 billion tonsof standard coal However because the unrenewablecharacteristics of the coal and the pollution causedby burning are bound to draw the attention of thecountry and the residents how to improve the coalutilization efficiency and promote the use of cleanenergy will become a main research direction in thefuture

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This research is supported by the National Natural ScienceFoundation of China (no 71103115) the China Postdoc-toral Science Foundation surface of funded projects (no2012M510580) the Fundamental Research Funds for theShaanxi province soft science (no 2012KRM95) The authorswould like to thank the anonymous referees

References

[1] L Suganthi and A A Samuel ldquoEnergy models for demandforecastingmdasha reviewrdquo Renewable and Sustainable EnergyReviews vol 16 no 2 pp 1223ndash1240 2012

[2] V S Ediger and S Akar ldquoARIMA forecasting of primary energydemand by fuel in Turkeyrdquo Energy Policy vol 35 no 3 pp 1701ndash1708 2007

[3] S HMohr and GM Evans ldquoForecasting coal production until2100rdquo Fuel vol 88 no 11 pp 2059ndash2067 2009

[4] A Unler ldquoImprovement of energy demand forecasts usingswarm intelligence the case of Turkey with projections to 2025rdquoEnergy Policy vol 36 no 6 pp 1937ndash1944 2008

10 Discrete Dynamics in Nature and Society

[5] C-C Lee and Y-B Chiu ldquoModeling OECD energy demand aninternational panel smooth transition error-correction modelrdquoInternational Review of Economics amp Finance vol 25 pp 372ndash383 2013

[6] H Chan and S Lee ldquoForecasting the demand for energy inChinardquo Energy Journal vol 17 no 1 pp 19ndash30 1996

[7] C V Hirschhausen and M Andres ldquoLong-term electricitydemand in Chinamdashfrom quantitative to qualitative growthrdquoEnergy Policy vol 28 no 4 pp 231ndash241 2000

[8] S-W Yu and K-J Zhu ldquoA hybrid procedure for energy demandforecasting in Chinardquo Energy vol 37 no 1 pp 396ndash404 2012

[9] S W Yu Y-M Wei and K Wang ldquoChinas primary energydemands in 2020 predictions from an MPSO-RBF estimationmodelrdquo Energy Conversion andManagement vol 61 pp 59ndash662012

[10] M Zhang H L Mu G Li and Y D Ning ldquoForecasting thetransport energy demand based on PLSR method in ChinardquoEnergy vol 34 no 9 pp 1396ndash1400 2009

[11] P Crompton and Y RWu ldquoEnergy consumption in China pasttrends and future directionsrdquo Energy Economics vol 27 no 1pp 195ndash208 2005

[12] J Chai Z Y Zhang J L Fu J E Guo and S Y WangldquoIdentification and analysis of the structural mutations in oilprice systemrdquo Journal of Management Sciences vol 27 no 2 pp133ndash144 2014 (Chinese)

[13] BM Francis L Moseley and S O Iyare ldquoEnergy consumptionand projected growth in selected Caribbean countriesrdquo EnergyEconomics vol 29 no 6 pp 1224ndash1232 2007

[14] L M Saini ldquoPeak load forecasting using Bayesian regulariza-tion Resilient and adaptive backpropagation learning basedartificial neural networksrdquo Electric Power Systems Research vol78 no 7 pp 1302ndash1310 2008

[15] P Lauret E Fock R N Randrianarivony and J-F Manicom-Ramsamy ldquoBayesian neural network approach to short timeload forecastingrdquo Energy Conversion and Management vol 49no 5 pp 1156ndash1166 2008

[16] C A Sims ldquoMacroeconomics and realityrdquo Econometrica vol48 no 1 pp 1ndash48 1980

[17] T Zha ldquoA dynamic multivariate model for use in formulatingpolicyrdquo Economic Review vol 83 no 1 pp 16ndash29 1998

[18] R B Litterman ldquoA Bayesian procedure for forecasting withvector autoregressionsrdquo Department of Economics WorkingPaper Massachusetts Institute of Technology 1980

[19] R B Litterman ldquoForecasting with Bayesian vector autoregres-sions five years of experiencerdquo Journal of Business and EconomicStatistics vol 4 no 1 pp 25ndash38 1986

[20] T Doan R B Litterman and C A Sims ldquoForecastingand conditional projection using realistic prior distributionsrdquoEconometric Reviews vol 3 no 1 pp 1ndash100 1984

[21] C A Sims ldquoA nine-variable probabilistic macroeconomic fore-casting modelrdquo in Business Cycles Indicators and Forecasting JH Stock and M W Watson Eds University of Chicago PressChicago Ill USA 1992

[22] C A Sims and T A Zha ldquoBayesian methods for dynamicmultivariate modelsrdquo International Economic Review vol 39no 4 pp 949ndash968 1998

[23] R J Hyndman and Y Khandakar ldquoAutomatic time seriesforecasting the forecast package for Rrdquo Journal of StatisticalSoftware vol 27 no 3 pp 1ndash22 2008

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical Problems in Engineering

Hindawi Publishing Corporationhttpwwwhindawicom

Differential EquationsInternational Journal of

Volume 2014

Applied MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical PhysicsAdvances in

Complex AnalysisJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OptimizationJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Operations ResearchAdvances in

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Function Spaces

Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of Mathematics and Mathematical Sciences

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Algebra

Discrete Dynamics in Nature and Society

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Decision SciencesAdvances in

Discrete MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Stochastic AnalysisInternational Journal of

6 Discrete Dynamics in Nature and Society

6000

8000

10000

12000

14000

16000

18000

050000

100000150000200000250000300000350000400000450000500000

1980

1985

1990

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

2018

2019

2020

Industry coal consumptionHousehold coal consumption

Coa

l con

sum

ptio

n (1

0000

tons

)

Coa

l con

sum

ptio

n (1

0000

tons

)

Figure 4 Changes of industry and household coal consumption

5 Structural Analysis of TotalCoal Consumption

A univariate ETS (exponential smoothing) forecast model isused to predict coal consumption of each sector in 2020 inthe forecast package for R Error Trend and Seasonal of ETSmodel denote error term (A M Z) trend term (N A M Z)and season term (NAM Z) respectively inwhich the groupincludes thirty models

From the coal balance sheet in China Statistical YearbookandWINDDatabase coal consumption data of seven sectorsare selected for the period 1980ndash2010 The balance equa-tion for coal consumption Total coal consumption (TO) =Agriculture Forestry Animal Husbandry Fishery andWaterConservancy (AFAFW) + Industry (IND) + Construction(CON) + Transport Storage and Post (TSP) +Wholesale andRetail Trades Hotels and Catering Services (WRHC) +OtherSectors (OS) + Household Consumption (HC) Unit is 10000tons

The finalmodels (Table 2) are selected from thirtymodelsby comparing information criterion (AIC BIC AICc) andforecast precision (mean error root mean square error meanabsolute error maximum permissible errors mean absolutepercentage error mean absolute error square) (Table 3)Command is run as follows R gt fcast lt -forecast(est(x))

As can be seen from the Table 4 the total coal consump-tion is the sum of coal consumption in seven sectors whichis 39728663 billion tons in 2015 This value is similar tothe goal in the development of coal industry in the twelfthfive-year plan which is 39 billion tons in 2015 By 2020Chinarsquos total coal consumption will reach 483 billion tonswith an annual growth rate of 436 of which industry andhousehold consumption account for the largest proportionIn Figure 4 the industry coal consumption was 438 milliontons 811 million tons 1278 billion tons and 296 billiontons in 1980 1990 2000 and 2010 respectively and it ispredicted to reach 4669 billion tons in 2020 accountingfor 9667 of total coal consumption Therefore to reducecoal consumption in China industrial energy conservationand emissions reduction are the fastest and most effective

breakthrough points including the adjustment of industrialstructure and optimization of technology

However as the second largest main channel the propor-tion of household consumption is decreasing year by yearFigure 4 shows that household consumption was up to 167million tons in 1990 which accounts for 1583 of the totalcoal consumption By 2020 household consumption willbe 092 million tons accounting for 189 of the total coalconsumption In the field of household energy consumptionas the diversity of the type of energy and development oftechnology electricity and heating oil are more available andconvenient than coal for residents so less and less coal is usedin the household rapidly This is beneficial to environmentalprotection and nonrenewable resources saving

As shown in Figure 5 transport storage and post sectorswere the largest coal consumption sectors in 1980 but theybecame the lowest coal consumption sectors after thirty yearsin 2010 and their coal consumption will decline constantlyduring the next decade by predictionsThemain reason is thatcoal-fired steam locomotive engines have been replaced byinternal combustion and electric locomotives by the railwaysSo most of the raw coal has been replaced by fuel andelectricity at the same time Nowadays road transporta-tion mainly consumes gasoline and diesel oil waterwaystransportation mainly consumes fuel oil and diesel oil airtransport mainly consumes aviation kerosene Thereforealthough the development of transport storage and postsectors still consumes large amounts of energy the usage ofcoal is insignificant

As the pillar industry of national economy in our countryconstruction sector is also an energy-intensive industryCoal consumption increased from 556 to 71892 milliontons between 1980 and 2010 representing a relatively smallgrowth By forecasting the amount of coal consumption isstill increasing in the future (Figure 5) At the same time theconstruction sector does not have a high proportion of coalconsumption and accounted for only 023 in 2010 That isbecause there are various types of energies consumed by theconstruction industry including raw coal gasoline diesel oilfuel oil heat electricity and other petroleum products With

Discrete Dynamics in Nature and Society 7

300

600

900

1200

1500

1800

2100

2400

2700

1980

1985

1990

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

2018

2019

2020

AFAFWCONTSP

WRHCOS

Coa

l con

sum

ptio

n (1

0000

tons

)

Figure 5 Changes of other sectorsrsquo coal consumption

Table 2 Formulae for recursive calculations point forecasts and parametric estimated value on sector by ETS method

Variables Forms Models and parameters

AFAFW (M N N)119897119905= 120572119910119905+ (1 minus 120572) 119897

119905minus1 119910119905+ℎ|119905

= 119897119905

119910119905+1

= 119910119905+1|119905

(1 + 120576119905) 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 00001 119897 = 17937859 120590 = 01809

IND (A A N)

119897119905= 120572119910119905+ (1 minus 120572) (119897

119905minus1+ 119887119905minus1) 119887119905= 120573 (119897

119905minus 119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905+ ℎ119887119905

119910119905+1

= 119910119905+1|119905

+ 120576119905 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 09999 120573 = 02697119897 = 547821109 119887 = 93385498 120590 = 129103600

CON (A A N)119897119905= 120572119910119905+ (1 minus 120572) (119897

119905minus1+ 119887119905minus1) 119887119905= 120573 (119897

119905minus 119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905+ ℎ119887119905

119910119905+1

= 119910119905+1|119905

+ 120576119905 120576119905sim 119873119868119863 (0 120590

2) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 00009 120573 = 00009 119897 = 4483868 119887 = 87382 120590 = 536530

TSP (M M N)119897119905= 119886119910119905+ (1 minus 119886) 119897

119905minus1119887119905minus1

119887119905= 120573 (119897

119905119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905119887119905

ℎ

119910119905+1

= 119910119905+1|119905

(1 + 120576119905) 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 00001 120573 = 00001 119897 = 23050531 119887 = 09371 120590 = 01049

WRHC (M M N)119897119905= 119886119910119905+ (1 minus 119886) 119897

119905minus1119887119905minus1

119887119905= 120573 (119897

119905119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905119887119905

ℎ

119910119905+1

= 119910119905+1|119905

(1 + 120576119905) 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 00001 120573 = 00001 119897 = 7651436 119887 = 10328 120590 = 02562

OS (A N N)119897119905= 119886119910119905+ (1 minus 119886) 119897

119905minus1 119910119905+ℎ|119905

= 119897119905

119910119905+1

= 119910119905+1|119905

+ 120576119905 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 08917 119897 = 11487978 120590 = 4978256

HC (M N N)119897119905= 119886119910119905+ (1 minus 119886) 119897

119905minus1 119910119905+ℎ|119905

= 119897119905

119910119905+1

= 119910119905+1|119905

(1 + 120576119905) 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 09999 119897 = 113736307 120590 = 01327

TO (A A N)119897119905= 120572119910119905+ (1 minus 120572) (119897

119905minus1+ 119887119905minus1) 119887119905= 120573 (119897

119905minus 119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905+ ℎ119887119905

119910119905+1

= 119910119905+1|119905

+ 120576119905 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 09999 120573 = 01665 119897 = 784219911 119887 = 93651674 120590 = 146462100

In the above formulae for recursive calculations and point forecasts 119897119905denotes the series level at time 119905 119887

119905denotes the

slope at time 119905 120572 and 120573 are smoothness index ℎ denotes the lag period

the development of technology the construction sector willbecome dependent on power and all kinds of oil productsincreasingly instead of coal which can improve the energyefficiency and control environmental pollution gradually

Wholesale and retail trade and hotels and catering ser-vices consume coal electricity heat gasoline diesel oil lique-fied petroleum gas natural gas and other varieties of energy

Influenced by price changes and economic development thissector has gone through booms and busts repeatedly From1980 to 1990 the sector had a rising trend due to adaptingto the market demand quickly from 1990 to 1997 with highprice the sector was in the integration stage from 1997 to1999 the sector was in a low price and low growth stagefrom 2000 to 2004 the sector was in a low price and high

8 Discrete Dynamics in Nature and Society

Table 3 Accuracy of forecast models on sector by ETS method

AFAFW IND CON TSP WRHC OS HC TOAIC 3107869 4694980 2392015 2724609 3085274 3287655 3715486 4747963AICc 3114536 4719980 2417015 2749609 3110274 3294321 3722153 4772963BIC 3128760 4736761 2433796 2766390 3127055 3308545 3736377 4789744ME minus01636 12971479 02783 53547 minus218557 455007 minus1054612 19036901RMSE 3244691 129103640 536530 1544569 3098943 4978256 15404774 146462092MAE 2556755 106357804 396210 1158266 2727594 3650402 10815817 118112234MPE minus40490 minus02898 minus10556 minus08693 minus105543 minus64377 minus19142 minus01609MAPE 157232 98982 77865 87576 271766 300246 96963 93149MASE 07907 07024 08702 07315 11268 09915 09609 07435

Table 4 Forecasting result of coal consumption on sector by ETS method

ETS TO AFAFW IND CON TSP WRHC OS HC2011 32882036 179379 31311760 64068 55260 155568 200086 9159172012 34593125 179379 33020350 64942 51787 160666 200086 9159172013 36304590 179379 34728930 65816 48532 165931 200086 9159172014 38016441 179379 36437520 66691 45481 171368 200086 9159172015 39728663 179379 38146110 67565 42623 176984 200086 9159172016 41441238 179379 39854690 68439 39944 182784 200086 9159172017 43154182 179379 41563280 69314 37433 188774 200086 9159172018 44867480 179379 43271870 70188 35080 194961 200086 9159172019 46581118 179379 44980450 71062 32875 201350 200086 9159172020 48295114 179379 46689040 71937 30809 207948 200086 915917Unit 10000 tons

growth stage domestic prices began to soar in 2005 andsince the American subprime crisis and the domestic policyadjustment in 2008 the sector appears to have had violentfluctuations However Figure 5 shows that the wholesale andretail trade hotels and catering services will have a growthtrend of coal consumption in the future Industry scale willcontinue to expand which is consistent with our countryrsquospolicy to develop the tertiary industry vigorously

Agriculture forestry animal husbandry fishery andwater conservancy are the foundation of the national econ-omy Coal consumption was in the second place before 1990but in 1993 2000 2005 and 2008 there was a great fall(Figure 5) in contrast coal consumption of wholesale andretail trade hotels and catering services and other sectorshad an obvious increase in these three years Considerablecontact with the change of the prices is guessed

6 Results and Discussion

Table 5 shows prediction results of BVAR models under thethree different prior distributions and the ETS models Theunit is 10000 tons in ETS models which is 10000 tons ofstandard coal in BVAR models So the results of ETS modelare multiplied by the converted coefficient of 07143 as tons ofstandard coal

The result of BVAR models under the normal-Wishartprior distribution (prior = 0) is the same as under the normal-flat prior distribution (prior = 1) Using the ETS model to

Table 5 Total coal consumption forecast

BVARPrior = 0 or 1

BVARPrior = 2

ETSaggregation

ETSforecast

2011 23803337 23803337 234876 2344752012 25268675 25403407 247099 2459212013 26815268 25766741 259324 2573662014 28449414 25711339 271551 2688112015 3017629 27462729 283782 2802562016 32001513 31472065 296014 2917012017 33931196 36070012 308250 3031462018 35972003 40119003 320489 3145912019 38131212 44491339 332729 3260372020 40416784 50482369 344972 337481Unit 10000 tons of standard coal

predict coal consumption of every sector and then addingthe seven values to get 344972 billion tons of standard coalin 2020 only based on the historical total coal consumption337481 billion tons of standard coal are obtained in 2020directly FromTable 5 in 2015 forecast results of fourmethodshave small differences But in 2020 the results of BVARmodelare larger than ETS model (Figure 6)

The reasonmay be that in BVARmodel the gross value ofindustrial output and the downstream industrial productionare influencing factors whose values are increasing over the

Discrete Dynamics in Nature and Society 9

20

25

30

35

40

45

50

55

2011 2012 2013 2014 2015 2016 2017 2018 2019 2020

BVAR 0 1BVAR 2

ETS AETS F

Coa

l con

sum

ptio

n (te

n th

ousa

nd to

ns o

f sta

ndar

d co

al)

times104

Figure 6 Total coal consumption forecast

years So they drive the rise in total coal consumption in thelong run However in ETS model total coal consumption isthe sum of coal consumption in seven sectors and the abovehas introduced that the coal consumption of some sectorshas showed a decreasing trend such as agriculture forestryanimal husbandry fishery and water conservancy transportstorage and post So the ETS model prediction results aresmaller

7 Conclusions

China is a big developing country in the process of industri-alization and urbanization and therefore coal consumptionwill remain high in the future for a long time Based on guar-anteeing the stability of the VAR system coal consumptiongross industrial output value and the downstream industrialproduction (crude steel cement thermal power) are selectedas variables to establish the BVAR (2) model At the sametime this paper analyses the changes of coal consumptionstructure based on sector division The forecast models areselected through comparing information criteria and forecastaccuracy comparison from thirty ETSmodels By forecastingthe total coal consumption and analyzing structural changesthis paper gets the following main conclusions

(1) Results of the impulse response function indicate thatthe increased output of the downstream industry willbecome a driving force of coal consumption In thefirst two periods the same response has been therewith the increase of the gross value of industrialoutput Later there will be a negative response

(2) The variance decomposition results show that thecontribution of gross value of industrial outputaccounted for 50of growth in coal consumption andhas long-term stable influence in the downstreamindustry the contribution rate of crude steel outputis greater than that of the cement output and thermalpower output

(3) In the next decade from a sector division point ofview industry and household consumption account

for the largest proportion of total coal consumptionthe proportion of industrial coal consumption willcontinue to rise but the proportion of householdcoal consumption will continue to decline Transportstorage and post sectors were the largest consumersof coal consumption in 1980 which will becomethe lowest in the future Construction sector despiteits lower reliance on coal has experienced a risein coal consumption in recent years The scales ofconsumption of wholesale and retail trade hotelsand catering services agriculture forestry animalhusbandry fishery and water conservancy sectorswere influenced by price fluctuations in the past fewdecades causing the fluctuation of coal consumptionBut in the next decade coal consumption of bothsectors will have a rising trend

(4) TheBVAR (2) forecastmodels predict that in 2015 coaldemand will reach 275ndash302 billion tons of standardcoal and in 2020 it will reach 404ndash505 billion tonsof standard coal in China That is to say as themomentumof rapid development of Chinarsquos economyin the future the coal consumption in 2020 is almostdouble that of in 2012 which is 246ndash253 billion tonsof standard coal However because the unrenewablecharacteristics of the coal and the pollution causedby burning are bound to draw the attention of thecountry and the residents how to improve the coalutilization efficiency and promote the use of cleanenergy will become a main research direction in thefuture

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This research is supported by the National Natural ScienceFoundation of China (no 71103115) the China Postdoc-toral Science Foundation surface of funded projects (no2012M510580) the Fundamental Research Funds for theShaanxi province soft science (no 2012KRM95) The authorswould like to thank the anonymous referees

References

[1] L Suganthi and A A Samuel ldquoEnergy models for demandforecastingmdasha reviewrdquo Renewable and Sustainable EnergyReviews vol 16 no 2 pp 1223ndash1240 2012

[2] V S Ediger and S Akar ldquoARIMA forecasting of primary energydemand by fuel in Turkeyrdquo Energy Policy vol 35 no 3 pp 1701ndash1708 2007

[3] S HMohr and GM Evans ldquoForecasting coal production until2100rdquo Fuel vol 88 no 11 pp 2059ndash2067 2009

[4] A Unler ldquoImprovement of energy demand forecasts usingswarm intelligence the case of Turkey with projections to 2025rdquoEnergy Policy vol 36 no 6 pp 1937ndash1944 2008

10 Discrete Dynamics in Nature and Society

[5] C-C Lee and Y-B Chiu ldquoModeling OECD energy demand aninternational panel smooth transition error-correction modelrdquoInternational Review of Economics amp Finance vol 25 pp 372ndash383 2013

[6] H Chan and S Lee ldquoForecasting the demand for energy inChinardquo Energy Journal vol 17 no 1 pp 19ndash30 1996

[7] C V Hirschhausen and M Andres ldquoLong-term electricitydemand in Chinamdashfrom quantitative to qualitative growthrdquoEnergy Policy vol 28 no 4 pp 231ndash241 2000

[8] S-W Yu and K-J Zhu ldquoA hybrid procedure for energy demandforecasting in Chinardquo Energy vol 37 no 1 pp 396ndash404 2012

[9] S W Yu Y-M Wei and K Wang ldquoChinas primary energydemands in 2020 predictions from an MPSO-RBF estimationmodelrdquo Energy Conversion andManagement vol 61 pp 59ndash662012

[10] M Zhang H L Mu G Li and Y D Ning ldquoForecasting thetransport energy demand based on PLSR method in ChinardquoEnergy vol 34 no 9 pp 1396ndash1400 2009

[11] P Crompton and Y RWu ldquoEnergy consumption in China pasttrends and future directionsrdquo Energy Economics vol 27 no 1pp 195ndash208 2005

[12] J Chai Z Y Zhang J L Fu J E Guo and S Y WangldquoIdentification and analysis of the structural mutations in oilprice systemrdquo Journal of Management Sciences vol 27 no 2 pp133ndash144 2014 (Chinese)

[13] BM Francis L Moseley and S O Iyare ldquoEnergy consumptionand projected growth in selected Caribbean countriesrdquo EnergyEconomics vol 29 no 6 pp 1224ndash1232 2007

[14] L M Saini ldquoPeak load forecasting using Bayesian regulariza-tion Resilient and adaptive backpropagation learning basedartificial neural networksrdquo Electric Power Systems Research vol78 no 7 pp 1302ndash1310 2008

[15] P Lauret E Fock R N Randrianarivony and J-F Manicom-Ramsamy ldquoBayesian neural network approach to short timeload forecastingrdquo Energy Conversion and Management vol 49no 5 pp 1156ndash1166 2008

[16] C A Sims ldquoMacroeconomics and realityrdquo Econometrica vol48 no 1 pp 1ndash48 1980

[17] T Zha ldquoA dynamic multivariate model for use in formulatingpolicyrdquo Economic Review vol 83 no 1 pp 16ndash29 1998

[18] R B Litterman ldquoA Bayesian procedure for forecasting withvector autoregressionsrdquo Department of Economics WorkingPaper Massachusetts Institute of Technology 1980

[19] R B Litterman ldquoForecasting with Bayesian vector autoregres-sions five years of experiencerdquo Journal of Business and EconomicStatistics vol 4 no 1 pp 25ndash38 1986

[20] T Doan R B Litterman and C A Sims ldquoForecastingand conditional projection using realistic prior distributionsrdquoEconometric Reviews vol 3 no 1 pp 1ndash100 1984

[21] C A Sims ldquoA nine-variable probabilistic macroeconomic fore-casting modelrdquo in Business Cycles Indicators and Forecasting JH Stock and M W Watson Eds University of Chicago PressChicago Ill USA 1992

[22] C A Sims and T A Zha ldquoBayesian methods for dynamicmultivariate modelsrdquo International Economic Review vol 39no 4 pp 949ndash968 1998

[23] R J Hyndman and Y Khandakar ldquoAutomatic time seriesforecasting the forecast package for Rrdquo Journal of StatisticalSoftware vol 27 no 3 pp 1ndash22 2008

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical Problems in Engineering

Hindawi Publishing Corporationhttpwwwhindawicom

Differential EquationsInternational Journal of

Volume 2014

Applied MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical PhysicsAdvances in

Complex AnalysisJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OptimizationJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Operations ResearchAdvances in

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Function Spaces

Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of Mathematics and Mathematical Sciences

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Algebra

Discrete Dynamics in Nature and Society

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Decision SciencesAdvances in

Discrete MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Stochastic AnalysisInternational Journal of

Discrete Dynamics in Nature and Society 7

300

600

900

1200

1500

1800

2100

2400

2700

1980

1985

1990

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

2018

2019

2020

AFAFWCONTSP

WRHCOS

Coa

l con

sum

ptio

n (1

0000

tons

)

Figure 5 Changes of other sectorsrsquo coal consumption

Table 2 Formulae for recursive calculations point forecasts and parametric estimated value on sector by ETS method

Variables Forms Models and parameters

AFAFW (M N N)119897119905= 120572119910119905+ (1 minus 120572) 119897

119905minus1 119910119905+ℎ|119905

= 119897119905

119910119905+1

= 119910119905+1|119905

(1 + 120576119905) 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 00001 119897 = 17937859 120590 = 01809

IND (A A N)

119897119905= 120572119910119905+ (1 minus 120572) (119897

119905minus1+ 119887119905minus1) 119887119905= 120573 (119897

119905minus 119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905+ ℎ119887119905

119910119905+1

= 119910119905+1|119905

+ 120576119905 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 09999 120573 = 02697119897 = 547821109 119887 = 93385498 120590 = 129103600

CON (A A N)119897119905= 120572119910119905+ (1 minus 120572) (119897

119905minus1+ 119887119905minus1) 119887119905= 120573 (119897

119905minus 119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905+ ℎ119887119905

119910119905+1

= 119910119905+1|119905

+ 120576119905 120576119905sim 119873119868119863 (0 120590

2) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 00009 120573 = 00009 119897 = 4483868 119887 = 87382 120590 = 536530

TSP (M M N)119897119905= 119886119910119905+ (1 minus 119886) 119897

119905minus1119887119905minus1

119887119905= 120573 (119897

119905119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905119887119905

ℎ

119910119905+1

= 119910119905+1|119905

(1 + 120576119905) 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 00001 120573 = 00001 119897 = 23050531 119887 = 09371 120590 = 01049

WRHC (M M N)119897119905= 119886119910119905+ (1 minus 119886) 119897

119905minus1119887119905minus1

119887119905= 120573 (119897

119905119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905119887119905

ℎ

119910119905+1

= 119910119905+1|119905

(1 + 120576119905) 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 00001 120573 = 00001 119897 = 7651436 119887 = 10328 120590 = 02562

OS (A N N)119897119905= 119886119910119905+ (1 minus 119886) 119897

119905minus1 119910119905+ℎ|119905

= 119897119905

119910119905+1

= 119910119905+1|119905

+ 120576119905 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 08917 119897 = 11487978 120590 = 4978256

HC (M N N)119897119905= 119886119910119905+ (1 minus 119886) 119897

119905minus1 119910119905+ℎ|119905

= 119897119905

119910119905+1

= 119910119905+1|119905

(1 + 120576119905) 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 09999 119897 = 113736307 120590 = 01327

TO (A A N)119897119905= 120572119910119905+ (1 minus 120572) (119897

119905minus1+ 119887119905minus1) 119887119905= 120573 (119897

119905minus 119897119905minus1) + (1 minus 120573) 119887

119905minus1 119910119905+ℎ|119905

= 119897119905+ ℎ119887119905

119910119905+1

= 119910119905+1|119905

+ 120576119905 120576119905sim 119873119868119863 (0 1205902) 0 lt 120572 lt 1 0 lt 120573 lt 120572

120572 = 09999 120573 = 01665 119897 = 784219911 119887 = 93651674 120590 = 146462100

In the above formulae for recursive calculations and point forecasts 119897119905denotes the series level at time 119905 119887

119905denotes the

slope at time 119905 120572 and 120573 are smoothness index ℎ denotes the lag period

the development of technology the construction sector willbecome dependent on power and all kinds of oil productsincreasingly instead of coal which can improve the energyefficiency and control environmental pollution gradually

Wholesale and retail trade and hotels and catering ser-vices consume coal electricity heat gasoline diesel oil lique-fied petroleum gas natural gas and other varieties of energy

Influenced by price changes and economic development thissector has gone through booms and busts repeatedly From1980 to 1990 the sector had a rising trend due to adaptingto the market demand quickly from 1990 to 1997 with highprice the sector was in the integration stage from 1997 to1999 the sector was in a low price and low growth stagefrom 2000 to 2004 the sector was in a low price and high

8 Discrete Dynamics in Nature and Society

Table 3 Accuracy of forecast models on sector by ETS method

AFAFW IND CON TSP WRHC OS HC TOAIC 3107869 4694980 2392015 2724609 3085274 3287655 3715486 4747963AICc 3114536 4719980 2417015 2749609 3110274 3294321 3722153 4772963BIC 3128760 4736761 2433796 2766390 3127055 3308545 3736377 4789744ME minus01636 12971479 02783 53547 minus218557 455007 minus1054612 19036901RMSE 3244691 129103640 536530 1544569 3098943 4978256 15404774 146462092MAE 2556755 106357804 396210 1158266 2727594 3650402 10815817 118112234MPE minus40490 minus02898 minus10556 minus08693 minus105543 minus64377 minus19142 minus01609MAPE 157232 98982 77865 87576 271766 300246 96963 93149MASE 07907 07024 08702 07315 11268 09915 09609 07435

Table 4 Forecasting result of coal consumption on sector by ETS method

ETS TO AFAFW IND CON TSP WRHC OS HC2011 32882036 179379 31311760 64068 55260 155568 200086 9159172012 34593125 179379 33020350 64942 51787 160666 200086 9159172013 36304590 179379 34728930 65816 48532 165931 200086 9159172014 38016441 179379 36437520 66691 45481 171368 200086 9159172015 39728663 179379 38146110 67565 42623 176984 200086 9159172016 41441238 179379 39854690 68439 39944 182784 200086 9159172017 43154182 179379 41563280 69314 37433 188774 200086 9159172018 44867480 179379 43271870 70188 35080 194961 200086 9159172019 46581118 179379 44980450 71062 32875 201350 200086 9159172020 48295114 179379 46689040 71937 30809 207948 200086 915917Unit 10000 tons

growth stage domestic prices began to soar in 2005 andsince the American subprime crisis and the domestic policyadjustment in 2008 the sector appears to have had violentfluctuations However Figure 5 shows that the wholesale andretail trade hotels and catering services will have a growthtrend of coal consumption in the future Industry scale willcontinue to expand which is consistent with our countryrsquospolicy to develop the tertiary industry vigorously

Agriculture forestry animal husbandry fishery andwater conservancy are the foundation of the national econ-omy Coal consumption was in the second place before 1990but in 1993 2000 2005 and 2008 there was a great fall(Figure 5) in contrast coal consumption of wholesale andretail trade hotels and catering services and other sectorshad an obvious increase in these three years Considerablecontact with the change of the prices is guessed

6 Results and Discussion

Table 5 shows prediction results of BVAR models under thethree different prior distributions and the ETS models Theunit is 10000 tons in ETS models which is 10000 tons ofstandard coal in BVAR models So the results of ETS modelare multiplied by the converted coefficient of 07143 as tons ofstandard coal

The result of BVAR models under the normal-Wishartprior distribution (prior = 0) is the same as under the normal-flat prior distribution (prior = 1) Using the ETS model to

Table 5 Total coal consumption forecast

BVARPrior = 0 or 1

BVARPrior = 2

ETSaggregation

ETSforecast

2011 23803337 23803337 234876 2344752012 25268675 25403407 247099 2459212013 26815268 25766741 259324 2573662014 28449414 25711339 271551 2688112015 3017629 27462729 283782 2802562016 32001513 31472065 296014 2917012017 33931196 36070012 308250 3031462018 35972003 40119003 320489 3145912019 38131212 44491339 332729 3260372020 40416784 50482369 344972 337481Unit 10000 tons of standard coal

predict coal consumption of every sector and then addingthe seven values to get 344972 billion tons of standard coalin 2020 only based on the historical total coal consumption337481 billion tons of standard coal are obtained in 2020directly FromTable 5 in 2015 forecast results of fourmethodshave small differences But in 2020 the results of BVARmodelare larger than ETS model (Figure 6)

The reasonmay be that in BVARmodel the gross value ofindustrial output and the downstream industrial productionare influencing factors whose values are increasing over the

Discrete Dynamics in Nature and Society 9

20

25

30

35

40

45

50

55

2011 2012 2013 2014 2015 2016 2017 2018 2019 2020

BVAR 0 1BVAR 2

ETS AETS F

Coa

l con

sum

ptio

n (te

n th

ousa

nd to

ns o

f sta

ndar

d co

al)

times104

Figure 6 Total coal consumption forecast

years So they drive the rise in total coal consumption in thelong run However in ETS model total coal consumption isthe sum of coal consumption in seven sectors and the abovehas introduced that the coal consumption of some sectorshas showed a decreasing trend such as agriculture forestryanimal husbandry fishery and water conservancy transportstorage and post So the ETS model prediction results aresmaller

7 Conclusions

China is a big developing country in the process of industri-alization and urbanization and therefore coal consumptionwill remain high in the future for a long time Based on guar-anteeing the stability of the VAR system coal consumptiongross industrial output value and the downstream industrialproduction (crude steel cement thermal power) are selectedas variables to establish the BVAR (2) model At the sametime this paper analyses the changes of coal consumptionstructure based on sector division The forecast models areselected through comparing information criteria and forecastaccuracy comparison from thirty ETSmodels By forecastingthe total coal consumption and analyzing structural changesthis paper gets the following main conclusions

(1) Results of the impulse response function indicate thatthe increased output of the downstream industry willbecome a driving force of coal consumption In thefirst two periods the same response has been therewith the increase of the gross value of industrialoutput Later there will be a negative response

(2) The variance decomposition results show that thecontribution of gross value of industrial outputaccounted for 50of growth in coal consumption andhas long-term stable influence in the downstreamindustry the contribution rate of crude steel outputis greater than that of the cement output and thermalpower output

(3) In the next decade from a sector division point ofview industry and household consumption account

for the largest proportion of total coal consumptionthe proportion of industrial coal consumption willcontinue to rise but the proportion of householdcoal consumption will continue to decline Transportstorage and post sectors were the largest consumersof coal consumption in 1980 which will becomethe lowest in the future Construction sector despiteits lower reliance on coal has experienced a risein coal consumption in recent years The scales ofconsumption of wholesale and retail trade hotelsand catering services agriculture forestry animalhusbandry fishery and water conservancy sectorswere influenced by price fluctuations in the past fewdecades causing the fluctuation of coal consumptionBut in the next decade coal consumption of bothsectors will have a rising trend

(4) TheBVAR (2) forecastmodels predict that in 2015 coaldemand will reach 275ndash302 billion tons of standardcoal and in 2020 it will reach 404ndash505 billion tonsof standard coal in China That is to say as themomentumof rapid development of Chinarsquos economyin the future the coal consumption in 2020 is almostdouble that of in 2012 which is 246ndash253 billion tonsof standard coal However because the unrenewablecharacteristics of the coal and the pollution causedby burning are bound to draw the attention of thecountry and the residents how to improve the coalutilization efficiency and promote the use of cleanenergy will become a main research direction in thefuture

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This research is supported by the National Natural ScienceFoundation of China (no 71103115) the China Postdoc-toral Science Foundation surface of funded projects (no2012M510580) the Fundamental Research Funds for theShaanxi province soft science (no 2012KRM95) The authorswould like to thank the anonymous referees

References

[1] L Suganthi and A A Samuel ldquoEnergy models for demandforecastingmdasha reviewrdquo Renewable and Sustainable EnergyReviews vol 16 no 2 pp 1223ndash1240 2012

[2] V S Ediger and S Akar ldquoARIMA forecasting of primary energydemand by fuel in Turkeyrdquo Energy Policy vol 35 no 3 pp 1701ndash1708 2007

[3] S HMohr and GM Evans ldquoForecasting coal production until2100rdquo Fuel vol 88 no 11 pp 2059ndash2067 2009

[4] A Unler ldquoImprovement of energy demand forecasts usingswarm intelligence the case of Turkey with projections to 2025rdquoEnergy Policy vol 36 no 6 pp 1937ndash1944 2008

10 Discrete Dynamics in Nature and Society

[5] C-C Lee and Y-B Chiu ldquoModeling OECD energy demand aninternational panel smooth transition error-correction modelrdquoInternational Review of Economics amp Finance vol 25 pp 372ndash383 2013

[6] H Chan and S Lee ldquoForecasting the demand for energy inChinardquo Energy Journal vol 17 no 1 pp 19ndash30 1996

[7] C V Hirschhausen and M Andres ldquoLong-term electricitydemand in Chinamdashfrom quantitative to qualitative growthrdquoEnergy Policy vol 28 no 4 pp 231ndash241 2000

[8] S-W Yu and K-J Zhu ldquoA hybrid procedure for energy demandforecasting in Chinardquo Energy vol 37 no 1 pp 396ndash404 2012

[9] S W Yu Y-M Wei and K Wang ldquoChinas primary energydemands in 2020 predictions from an MPSO-RBF estimationmodelrdquo Energy Conversion andManagement vol 61 pp 59ndash662012

[10] M Zhang H L Mu G Li and Y D Ning ldquoForecasting thetransport energy demand based on PLSR method in ChinardquoEnergy vol 34 no 9 pp 1396ndash1400 2009

[11] P Crompton and Y RWu ldquoEnergy consumption in China pasttrends and future directionsrdquo Energy Economics vol 27 no 1pp 195ndash208 2005

[12] J Chai Z Y Zhang J L Fu J E Guo and S Y WangldquoIdentification and analysis of the structural mutations in oilprice systemrdquo Journal of Management Sciences vol 27 no 2 pp133ndash144 2014 (Chinese)

[13] BM Francis L Moseley and S O Iyare ldquoEnergy consumptionand projected growth in selected Caribbean countriesrdquo EnergyEconomics vol 29 no 6 pp 1224ndash1232 2007

[14] L M Saini ldquoPeak load forecasting using Bayesian regulariza-tion Resilient and adaptive backpropagation learning basedartificial neural networksrdquo Electric Power Systems Research vol78 no 7 pp 1302ndash1310 2008

[15] P Lauret E Fock R N Randrianarivony and J-F Manicom-Ramsamy ldquoBayesian neural network approach to short timeload forecastingrdquo Energy Conversion and Management vol 49no 5 pp 1156ndash1166 2008

[16] C A Sims ldquoMacroeconomics and realityrdquo Econometrica vol48 no 1 pp 1ndash48 1980

[17] T Zha ldquoA dynamic multivariate model for use in formulatingpolicyrdquo Economic Review vol 83 no 1 pp 16ndash29 1998

[18] R B Litterman ldquoA Bayesian procedure for forecasting withvector autoregressionsrdquo Department of Economics WorkingPaper Massachusetts Institute of Technology 1980

[19] R B Litterman ldquoForecasting with Bayesian vector autoregres-sions five years of experiencerdquo Journal of Business and EconomicStatistics vol 4 no 1 pp 25ndash38 1986

[20] T Doan R B Litterman and C A Sims ldquoForecastingand conditional projection using realistic prior distributionsrdquoEconometric Reviews vol 3 no 1 pp 1ndash100 1984

[21] C A Sims ldquoA nine-variable probabilistic macroeconomic fore-casting modelrdquo in Business Cycles Indicators and Forecasting JH Stock and M W Watson Eds University of Chicago PressChicago Ill USA 1992

[22] C A Sims and T A Zha ldquoBayesian methods for dynamicmultivariate modelsrdquo International Economic Review vol 39no 4 pp 949ndash968 1998

[23] R J Hyndman and Y Khandakar ldquoAutomatic time seriesforecasting the forecast package for Rrdquo Journal of StatisticalSoftware vol 27 no 3 pp 1ndash22 2008

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical Problems in Engineering

Hindawi Publishing Corporationhttpwwwhindawicom

Differential EquationsInternational Journal of

Volume 2014

Applied MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical PhysicsAdvances in

Complex AnalysisJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OptimizationJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Operations ResearchAdvances in

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Function Spaces

Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of Mathematics and Mathematical Sciences

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Algebra

Discrete Dynamics in Nature and Society

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Decision SciencesAdvances in

Discrete MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Stochastic AnalysisInternational Journal of

8 Discrete Dynamics in Nature and Society

Table 3 Accuracy of forecast models on sector by ETS method

AFAFW IND CON TSP WRHC OS HC TOAIC 3107869 4694980 2392015 2724609 3085274 3287655 3715486 4747963AICc 3114536 4719980 2417015 2749609 3110274 3294321 3722153 4772963BIC 3128760 4736761 2433796 2766390 3127055 3308545 3736377 4789744ME minus01636 12971479 02783 53547 minus218557 455007 minus1054612 19036901RMSE 3244691 129103640 536530 1544569 3098943 4978256 15404774 146462092MAE 2556755 106357804 396210 1158266 2727594 3650402 10815817 118112234MPE minus40490 minus02898 minus10556 minus08693 minus105543 minus64377 minus19142 minus01609MAPE 157232 98982 77865 87576 271766 300246 96963 93149MASE 07907 07024 08702 07315 11268 09915 09609 07435

Table 4 Forecasting result of coal consumption on sector by ETS method

ETS TO AFAFW IND CON TSP WRHC OS HC2011 32882036 179379 31311760 64068 55260 155568 200086 9159172012 34593125 179379 33020350 64942 51787 160666 200086 9159172013 36304590 179379 34728930 65816 48532 165931 200086 9159172014 38016441 179379 36437520 66691 45481 171368 200086 9159172015 39728663 179379 38146110 67565 42623 176984 200086 9159172016 41441238 179379 39854690 68439 39944 182784 200086 9159172017 43154182 179379 41563280 69314 37433 188774 200086 9159172018 44867480 179379 43271870 70188 35080 194961 200086 9159172019 46581118 179379 44980450 71062 32875 201350 200086 9159172020 48295114 179379 46689040 71937 30809 207948 200086 915917Unit 10000 tons

growth stage domestic prices began to soar in 2005 andsince the American subprime crisis and the domestic policyadjustment in 2008 the sector appears to have had violentfluctuations However Figure 5 shows that the wholesale andretail trade hotels and catering services will have a growthtrend of coal consumption in the future Industry scale willcontinue to expand which is consistent with our countryrsquospolicy to develop the tertiary industry vigorously

Agriculture forestry animal husbandry fishery andwater conservancy are the foundation of the national econ-omy Coal consumption was in the second place before 1990but in 1993 2000 2005 and 2008 there was a great fall(Figure 5) in contrast coal consumption of wholesale andretail trade hotels and catering services and other sectorshad an obvious increase in these three years Considerablecontact with the change of the prices is guessed

6 Results and Discussion

Table 5 shows prediction results of BVAR models under thethree different prior distributions and the ETS models Theunit is 10000 tons in ETS models which is 10000 tons ofstandard coal in BVAR models So the results of ETS modelare multiplied by the converted coefficient of 07143 as tons ofstandard coal

The result of BVAR models under the normal-Wishartprior distribution (prior = 0) is the same as under the normal-flat prior distribution (prior = 1) Using the ETS model to

Table 5 Total coal consumption forecast

BVARPrior = 0 or 1

BVARPrior = 2

ETSaggregation

ETSforecast

2011 23803337 23803337 234876 2344752012 25268675 25403407 247099 2459212013 26815268 25766741 259324 2573662014 28449414 25711339 271551 2688112015 3017629 27462729 283782 2802562016 32001513 31472065 296014 2917012017 33931196 36070012 308250 3031462018 35972003 40119003 320489 3145912019 38131212 44491339 332729 3260372020 40416784 50482369 344972 337481Unit 10000 tons of standard coal

predict coal consumption of every sector and then addingthe seven values to get 344972 billion tons of standard coalin 2020 only based on the historical total coal consumption337481 billion tons of standard coal are obtained in 2020directly FromTable 5 in 2015 forecast results of fourmethodshave small differences But in 2020 the results of BVARmodelare larger than ETS model (Figure 6)

The reasonmay be that in BVARmodel the gross value ofindustrial output and the downstream industrial productionare influencing factors whose values are increasing over the

Discrete Dynamics in Nature and Society 9

20

25

30

35

40

45

50

55

2011 2012 2013 2014 2015 2016 2017 2018 2019 2020

BVAR 0 1BVAR 2

ETS AETS F

Coa

l con

sum

ptio

n (te

n th

ousa

nd to

ns o

f sta

ndar

d co

al)

times104

Figure 6 Total coal consumption forecast

years So they drive the rise in total coal consumption in thelong run However in ETS model total coal consumption isthe sum of coal consumption in seven sectors and the abovehas introduced that the coal consumption of some sectorshas showed a decreasing trend such as agriculture forestryanimal husbandry fishery and water conservancy transportstorage and post So the ETS model prediction results aresmaller

7 Conclusions

China is a big developing country in the process of industri-alization and urbanization and therefore coal consumptionwill remain high in the future for a long time Based on guar-anteeing the stability of the VAR system coal consumptiongross industrial output value and the downstream industrialproduction (crude steel cement thermal power) are selectedas variables to establish the BVAR (2) model At the sametime this paper analyses the changes of coal consumptionstructure based on sector division The forecast models areselected through comparing information criteria and forecastaccuracy comparison from thirty ETSmodels By forecastingthe total coal consumption and analyzing structural changesthis paper gets the following main conclusions

(1) Results of the impulse response function indicate thatthe increased output of the downstream industry willbecome a driving force of coal consumption In thefirst two periods the same response has been therewith the increase of the gross value of industrialoutput Later there will be a negative response

(2) The variance decomposition results show that thecontribution of gross value of industrial outputaccounted for 50of growth in coal consumption andhas long-term stable influence in the downstreamindustry the contribution rate of crude steel outputis greater than that of the cement output and thermalpower output

(3) In the next decade from a sector division point ofview industry and household consumption account

for the largest proportion of total coal consumptionthe proportion of industrial coal consumption willcontinue to rise but the proportion of householdcoal consumption will continue to decline Transportstorage and post sectors were the largest consumersof coal consumption in 1980 which will becomethe lowest in the future Construction sector despiteits lower reliance on coal has experienced a risein coal consumption in recent years The scales ofconsumption of wholesale and retail trade hotelsand catering services agriculture forestry animalhusbandry fishery and water conservancy sectorswere influenced by price fluctuations in the past fewdecades causing the fluctuation of coal consumptionBut in the next decade coal consumption of bothsectors will have a rising trend

(4) TheBVAR (2) forecastmodels predict that in 2015 coaldemand will reach 275ndash302 billion tons of standardcoal and in 2020 it will reach 404ndash505 billion tonsof standard coal in China That is to say as themomentumof rapid development of Chinarsquos economyin the future the coal consumption in 2020 is almostdouble that of in 2012 which is 246ndash253 billion tonsof standard coal However because the unrenewablecharacteristics of the coal and the pollution causedby burning are bound to draw the attention of thecountry and the residents how to improve the coalutilization efficiency and promote the use of cleanenergy will become a main research direction in thefuture

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This research is supported by the National Natural ScienceFoundation of China (no 71103115) the China Postdoc-toral Science Foundation surface of funded projects (no2012M510580) the Fundamental Research Funds for theShaanxi province soft science (no 2012KRM95) The authorswould like to thank the anonymous referees

References

[1] L Suganthi and A A Samuel ldquoEnergy models for demandforecastingmdasha reviewrdquo Renewable and Sustainable EnergyReviews vol 16 no 2 pp 1223ndash1240 2012

[2] V S Ediger and S Akar ldquoARIMA forecasting of primary energydemand by fuel in Turkeyrdquo Energy Policy vol 35 no 3 pp 1701ndash1708 2007

[3] S HMohr and GM Evans ldquoForecasting coal production until2100rdquo Fuel vol 88 no 11 pp 2059ndash2067 2009

[4] A Unler ldquoImprovement of energy demand forecasts usingswarm intelligence the case of Turkey with projections to 2025rdquoEnergy Policy vol 36 no 6 pp 1937ndash1944 2008

10 Discrete Dynamics in Nature and Society

[5] C-C Lee and Y-B Chiu ldquoModeling OECD energy demand aninternational panel smooth transition error-correction modelrdquoInternational Review of Economics amp Finance vol 25 pp 372ndash383 2013

[6] H Chan and S Lee ldquoForecasting the demand for energy inChinardquo Energy Journal vol 17 no 1 pp 19ndash30 1996

[7] C V Hirschhausen and M Andres ldquoLong-term electricitydemand in Chinamdashfrom quantitative to qualitative growthrdquoEnergy Policy vol 28 no 4 pp 231ndash241 2000

[8] S-W Yu and K-J Zhu ldquoA hybrid procedure for energy demandforecasting in Chinardquo Energy vol 37 no 1 pp 396ndash404 2012

[9] S W Yu Y-M Wei and K Wang ldquoChinas primary energydemands in 2020 predictions from an MPSO-RBF estimationmodelrdquo Energy Conversion andManagement vol 61 pp 59ndash662012

[10] M Zhang H L Mu G Li and Y D Ning ldquoForecasting thetransport energy demand based on PLSR method in ChinardquoEnergy vol 34 no 9 pp 1396ndash1400 2009

[11] P Crompton and Y RWu ldquoEnergy consumption in China pasttrends and future directionsrdquo Energy Economics vol 27 no 1pp 195ndash208 2005

[12] J Chai Z Y Zhang J L Fu J E Guo and S Y WangldquoIdentification and analysis of the structural mutations in oilprice systemrdquo Journal of Management Sciences vol 27 no 2 pp133ndash144 2014 (Chinese)

[13] BM Francis L Moseley and S O Iyare ldquoEnergy consumptionand projected growth in selected Caribbean countriesrdquo EnergyEconomics vol 29 no 6 pp 1224ndash1232 2007

[14] L M Saini ldquoPeak load forecasting using Bayesian regulariza-tion Resilient and adaptive backpropagation learning basedartificial neural networksrdquo Electric Power Systems Research vol78 no 7 pp 1302ndash1310 2008

[15] P Lauret E Fock R N Randrianarivony and J-F Manicom-Ramsamy ldquoBayesian neural network approach to short timeload forecastingrdquo Energy Conversion and Management vol 49no 5 pp 1156ndash1166 2008

[16] C A Sims ldquoMacroeconomics and realityrdquo Econometrica vol48 no 1 pp 1ndash48 1980

[17] T Zha ldquoA dynamic multivariate model for use in formulatingpolicyrdquo Economic Review vol 83 no 1 pp 16ndash29 1998

[18] R B Litterman ldquoA Bayesian procedure for forecasting withvector autoregressionsrdquo Department of Economics WorkingPaper Massachusetts Institute of Technology 1980

[19] R B Litterman ldquoForecasting with Bayesian vector autoregres-sions five years of experiencerdquo Journal of Business and EconomicStatistics vol 4 no 1 pp 25ndash38 1986

[20] T Doan R B Litterman and C A Sims ldquoForecastingand conditional projection using realistic prior distributionsrdquoEconometric Reviews vol 3 no 1 pp 1ndash100 1984

[21] C A Sims ldquoA nine-variable probabilistic macroeconomic fore-casting modelrdquo in Business Cycles Indicators and Forecasting JH Stock and M W Watson Eds University of Chicago PressChicago Ill USA 1992

[22] C A Sims and T A Zha ldquoBayesian methods for dynamicmultivariate modelsrdquo International Economic Review vol 39no 4 pp 949ndash968 1998

[23] R J Hyndman and Y Khandakar ldquoAutomatic time seriesforecasting the forecast package for Rrdquo Journal of StatisticalSoftware vol 27 no 3 pp 1ndash22 2008

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical Problems in Engineering

Hindawi Publishing Corporationhttpwwwhindawicom

Differential EquationsInternational Journal of

Volume 2014

Applied MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical PhysicsAdvances in

Complex AnalysisJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OptimizationJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Operations ResearchAdvances in

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Function Spaces

Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of Mathematics and Mathematical Sciences

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Algebra

Discrete Dynamics in Nature and Society

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Decision SciencesAdvances in

Discrete MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Stochastic AnalysisInternational Journal of

Discrete Dynamics in Nature and Society 9

20

25

30

35

40

45

50

55

2011 2012 2013 2014 2015 2016 2017 2018 2019 2020

BVAR 0 1BVAR 2

ETS AETS F

Coa

l con

sum

ptio

n (te

n th

ousa

nd to

ns o

f sta

ndar

d co

al)

times104

Figure 6 Total coal consumption forecast

years So they drive the rise in total coal consumption in thelong run However in ETS model total coal consumption isthe sum of coal consumption in seven sectors and the abovehas introduced that the coal consumption of some sectorshas showed a decreasing trend such as agriculture forestryanimal husbandry fishery and water conservancy transportstorage and post So the ETS model prediction results aresmaller

7 Conclusions

China is a big developing country in the process of industri-alization and urbanization and therefore coal consumptionwill remain high in the future for a long time Based on guar-anteeing the stability of the VAR system coal consumptiongross industrial output value and the downstream industrialproduction (crude steel cement thermal power) are selectedas variables to establish the BVAR (2) model At the sametime this paper analyses the changes of coal consumptionstructure based on sector division The forecast models areselected through comparing information criteria and forecastaccuracy comparison from thirty ETSmodels By forecastingthe total coal consumption and analyzing structural changesthis paper gets the following main conclusions

(1) Results of the impulse response function indicate thatthe increased output of the downstream industry willbecome a driving force of coal consumption In thefirst two periods the same response has been therewith the increase of the gross value of industrialoutput Later there will be a negative response

(2) The variance decomposition results show that thecontribution of gross value of industrial outputaccounted for 50of growth in coal consumption andhas long-term stable influence in the downstreamindustry the contribution rate of crude steel outputis greater than that of the cement output and thermalpower output

(3) In the next decade from a sector division point ofview industry and household consumption account

for the largest proportion of total coal consumptionthe proportion of industrial coal consumption willcontinue to rise but the proportion of householdcoal consumption will continue to decline Transportstorage and post sectors were the largest consumersof coal consumption in 1980 which will becomethe lowest in the future Construction sector despiteits lower reliance on coal has experienced a risein coal consumption in recent years The scales ofconsumption of wholesale and retail trade hotelsand catering services agriculture forestry animalhusbandry fishery and water conservancy sectorswere influenced by price fluctuations in the past fewdecades causing the fluctuation of coal consumptionBut in the next decade coal consumption of bothsectors will have a rising trend

(4) TheBVAR (2) forecastmodels predict that in 2015 coaldemand will reach 275ndash302 billion tons of standardcoal and in 2020 it will reach 404ndash505 billion tonsof standard coal in China That is to say as themomentumof rapid development of Chinarsquos economyin the future the coal consumption in 2020 is almostdouble that of in 2012 which is 246ndash253 billion tonsof standard coal However because the unrenewablecharacteristics of the coal and the pollution causedby burning are bound to draw the attention of thecountry and the residents how to improve the coalutilization efficiency and promote the use of cleanenergy will become a main research direction in thefuture

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This research is supported by the National Natural ScienceFoundation of China (no 71103115) the China Postdoc-toral Science Foundation surface of funded projects (no2012M510580) the Fundamental Research Funds for theShaanxi province soft science (no 2012KRM95) The authorswould like to thank the anonymous referees

References

[1] L Suganthi and A A Samuel ldquoEnergy models for demandforecastingmdasha reviewrdquo Renewable and Sustainable EnergyReviews vol 16 no 2 pp 1223ndash1240 2012

[2] V S Ediger and S Akar ldquoARIMA forecasting of primary energydemand by fuel in Turkeyrdquo Energy Policy vol 35 no 3 pp 1701ndash1708 2007

[3] S HMohr and GM Evans ldquoForecasting coal production until2100rdquo Fuel vol 88 no 11 pp 2059ndash2067 2009

[4] A Unler ldquoImprovement of energy demand forecasts usingswarm intelligence the case of Turkey with projections to 2025rdquoEnergy Policy vol 36 no 6 pp 1937ndash1944 2008

10 Discrete Dynamics in Nature and Society

[5] C-C Lee and Y-B Chiu ldquoModeling OECD energy demand aninternational panel smooth transition error-correction modelrdquoInternational Review of Economics amp Finance vol 25 pp 372ndash383 2013

[6] H Chan and S Lee ldquoForecasting the demand for energy inChinardquo Energy Journal vol 17 no 1 pp 19ndash30 1996

[7] C V Hirschhausen and M Andres ldquoLong-term electricitydemand in Chinamdashfrom quantitative to qualitative growthrdquoEnergy Policy vol 28 no 4 pp 231ndash241 2000

[8] S-W Yu and K-J Zhu ldquoA hybrid procedure for energy demandforecasting in Chinardquo Energy vol 37 no 1 pp 396ndash404 2012

[9] S W Yu Y-M Wei and K Wang ldquoChinas primary energydemands in 2020 predictions from an MPSO-RBF estimationmodelrdquo Energy Conversion andManagement vol 61 pp 59ndash662012

[10] M Zhang H L Mu G Li and Y D Ning ldquoForecasting thetransport energy demand based on PLSR method in ChinardquoEnergy vol 34 no 9 pp 1396ndash1400 2009

[11] P Crompton and Y RWu ldquoEnergy consumption in China pasttrends and future directionsrdquo Energy Economics vol 27 no 1pp 195ndash208 2005

[12] J Chai Z Y Zhang J L Fu J E Guo and S Y WangldquoIdentification and analysis of the structural mutations in oilprice systemrdquo Journal of Management Sciences vol 27 no 2 pp133ndash144 2014 (Chinese)

[13] BM Francis L Moseley and S O Iyare ldquoEnergy consumptionand projected growth in selected Caribbean countriesrdquo EnergyEconomics vol 29 no 6 pp 1224ndash1232 2007

[14] L M Saini ldquoPeak load forecasting using Bayesian regulariza-tion Resilient and adaptive backpropagation learning basedartificial neural networksrdquo Electric Power Systems Research vol78 no 7 pp 1302ndash1310 2008

[15] P Lauret E Fock R N Randrianarivony and J-F Manicom-Ramsamy ldquoBayesian neural network approach to short timeload forecastingrdquo Energy Conversion and Management vol 49no 5 pp 1156ndash1166 2008

[16] C A Sims ldquoMacroeconomics and realityrdquo Econometrica vol48 no 1 pp 1ndash48 1980

[17] T Zha ldquoA dynamic multivariate model for use in formulatingpolicyrdquo Economic Review vol 83 no 1 pp 16ndash29 1998

[18] R B Litterman ldquoA Bayesian procedure for forecasting withvector autoregressionsrdquo Department of Economics WorkingPaper Massachusetts Institute of Technology 1980

[19] R B Litterman ldquoForecasting with Bayesian vector autoregres-sions five years of experiencerdquo Journal of Business and EconomicStatistics vol 4 no 1 pp 25ndash38 1986

[20] T Doan R B Litterman and C A Sims ldquoForecastingand conditional projection using realistic prior distributionsrdquoEconometric Reviews vol 3 no 1 pp 1ndash100 1984

[21] C A Sims ldquoA nine-variable probabilistic macroeconomic fore-casting modelrdquo in Business Cycles Indicators and Forecasting JH Stock and M W Watson Eds University of Chicago PressChicago Ill USA 1992

[22] C A Sims and T A Zha ldquoBayesian methods for dynamicmultivariate modelsrdquo International Economic Review vol 39no 4 pp 949ndash968 1998

[23] R J Hyndman and Y Khandakar ldquoAutomatic time seriesforecasting the forecast package for Rrdquo Journal of StatisticalSoftware vol 27 no 3 pp 1ndash22 2008

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical Problems in Engineering

Hindawi Publishing Corporationhttpwwwhindawicom

Differential EquationsInternational Journal of

Volume 2014

Applied MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical PhysicsAdvances in

Complex AnalysisJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OptimizationJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Operations ResearchAdvances in

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Function Spaces

Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of Mathematics and Mathematical Sciences

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Algebra

Discrete Dynamics in Nature and Society

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Decision SciencesAdvances in

Discrete MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Stochastic AnalysisInternational Journal of

10 Discrete Dynamics in Nature and Society

[5] C-C Lee and Y-B Chiu ldquoModeling OECD energy demand aninternational panel smooth transition error-correction modelrdquoInternational Review of Economics amp Finance vol 25 pp 372ndash383 2013

[6] H Chan and S Lee ldquoForecasting the demand for energy inChinardquo Energy Journal vol 17 no 1 pp 19ndash30 1996

[7] C V Hirschhausen and M Andres ldquoLong-term electricitydemand in Chinamdashfrom quantitative to qualitative growthrdquoEnergy Policy vol 28 no 4 pp 231ndash241 2000

[8] S-W Yu and K-J Zhu ldquoA hybrid procedure for energy demandforecasting in Chinardquo Energy vol 37 no 1 pp 396ndash404 2012

[9] S W Yu Y-M Wei and K Wang ldquoChinas primary energydemands in 2020 predictions from an MPSO-RBF estimationmodelrdquo Energy Conversion andManagement vol 61 pp 59ndash662012

[10] M Zhang H L Mu G Li and Y D Ning ldquoForecasting thetransport energy demand based on PLSR method in ChinardquoEnergy vol 34 no 9 pp 1396ndash1400 2009

[11] P Crompton and Y RWu ldquoEnergy consumption in China pasttrends and future directionsrdquo Energy Economics vol 27 no 1pp 195ndash208 2005

[12] J Chai Z Y Zhang J L Fu J E Guo and S Y WangldquoIdentification and analysis of the structural mutations in oilprice systemrdquo Journal of Management Sciences vol 27 no 2 pp133ndash144 2014 (Chinese)

[13] BM Francis L Moseley and S O Iyare ldquoEnergy consumptionand projected growth in selected Caribbean countriesrdquo EnergyEconomics vol 29 no 6 pp 1224ndash1232 2007

[14] L M Saini ldquoPeak load forecasting using Bayesian regulariza-tion Resilient and adaptive backpropagation learning basedartificial neural networksrdquo Electric Power Systems Research vol78 no 7 pp 1302ndash1310 2008

[15] P Lauret E Fock R N Randrianarivony and J-F Manicom-Ramsamy ldquoBayesian neural network approach to short timeload forecastingrdquo Energy Conversion and Management vol 49no 5 pp 1156ndash1166 2008

[16] C A Sims ldquoMacroeconomics and realityrdquo Econometrica vol48 no 1 pp 1ndash48 1980

[17] T Zha ldquoA dynamic multivariate model for use in formulatingpolicyrdquo Economic Review vol 83 no 1 pp 16ndash29 1998

[18] R B Litterman ldquoA Bayesian procedure for forecasting withvector autoregressionsrdquo Department of Economics WorkingPaper Massachusetts Institute of Technology 1980

[19] R B Litterman ldquoForecasting with Bayesian vector autoregres-sions five years of experiencerdquo Journal of Business and EconomicStatistics vol 4 no 1 pp 25ndash38 1986

[20] T Doan R B Litterman and C A Sims ldquoForecastingand conditional projection using realistic prior distributionsrdquoEconometric Reviews vol 3 no 1 pp 1ndash100 1984

[21] C A Sims ldquoA nine-variable probabilistic macroeconomic fore-casting modelrdquo in Business Cycles Indicators and Forecasting JH Stock and M W Watson Eds University of Chicago PressChicago Ill USA 1992

[22] C A Sims and T A Zha ldquoBayesian methods for dynamicmultivariate modelsrdquo International Economic Review vol 39no 4 pp 949ndash968 1998

[23] R J Hyndman and Y Khandakar ldquoAutomatic time seriesforecasting the forecast package for Rrdquo Journal of StatisticalSoftware vol 27 no 3 pp 1ndash22 2008

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical Problems in Engineering

Hindawi Publishing Corporationhttpwwwhindawicom

Differential EquationsInternational Journal of

Volume 2014

Applied MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical PhysicsAdvances in

Complex AnalysisJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OptimizationJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Operations ResearchAdvances in

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Function Spaces

Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of Mathematics and Mathematical Sciences

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Algebra

Discrete Dynamics in Nature and Society

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Decision SciencesAdvances in

Discrete MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Stochastic AnalysisInternational Journal of

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical Problems in Engineering

Hindawi Publishing Corporationhttpwwwhindawicom

Differential EquationsInternational Journal of

Volume 2014

Applied MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical PhysicsAdvances in

Complex AnalysisJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OptimizationJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Operations ResearchAdvances in

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Function Spaces

Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of Mathematics and Mathematical Sciences

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Algebra

Discrete Dynamics in Nature and Society

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Decision SciencesAdvances in

Discrete MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Stochastic AnalysisInternational Journal of