lable at ScienceDirect

Energy 87 (2015) 566e577

Contents lists avai

Energy

journal homepage: www.elsevier .com/locate/energy

Online hydraulic calculation and operation optimization of industrialsteam heating networks considering heat dissipation in pipes

Wei Zhong a, *, Hongcui Feng a, Xuguang Wang a, Dingfei Wu b, Minghua Xue c,Jian Wang c

a Institute of Thermal Science and Power System, Zhejiang University, Hangzhou, 310027, Chinab Caojing Cogeneration CO., LTD., Shanghai, 201507, Chinac Minghua Electric Power Technology and Engineering CO., LTD., Shanghai, 200437, China

a r t i c l e i n f o

Article history:Received 15 December 2014Received in revised form30 April 2015Accepted 5 May 2015Available online 29 May 2015

Keywords:Steam heating networkCIWH (condensation-induced waterhammer)Flow regimeHeat dissipationOnline hydraulic calculationOperation optimization

* Corresponding author. Tel.: þ86 13989882228; faE-mail address: wzhong@zju.edu.cn (W. Zhong).

http://dx.doi.org/10.1016/j.energy.2015.05.0240360-5442/© 2015 Elsevier Ltd. All rights reserved.

a b s t r a c t

Constructing industrial parks with DH (district heating) systems has become a main way to developmodern industry, which requires strict security and reliability of heating networks. Industrial steamheating networks are usually ring-shaped with multiple heating sources, and the working conditionswould be changeable due to the high frequency and a wide range of load variation of heating con-sumers. Under a specific working condition, low steam velocity for a long time (namely “steamstagnation”) in certain pipes will result in CIWH (condensation-induced water hammer) which willthreaten the security of the whole DH system. In this paper, a hydraulic calculation model is built tostudy the steam flow regime considering heat dissipation and condensation in pipes, an operationoptimization method is proposed to help eliminate steam stagnation through optimizing the heat loaddistribution of each heating source, a general software system entitled “HEATNET” is presented torealize online hydraulic calculation and operation optimization for arbitrary structured heating net-works. The practical application of HEATNET in Shanghai chemical industry zone shows that heatdissipation and condensation in pipes would influence the overall hydraulic calculation of steamheating networks and it can prevent CIWH and improve the security and reliability of steam heatingnetworks.

© 2015 Elsevier Ltd. All rights reserved.

1. Introduction

In recent years, constructing centralized industrial parks withDH (district heating) systems for specific industries has become animportant development pattern in China [1]. Replacing self-builtboiler rooms with DH systems in industrial parks is of great sig-nificance in improving the reliability and economy of heating sys-tems [2,3].

What's more, a substantial reduction in fuel demands and CO2emissions as well as cost can be achieved, and it is also easy forpollutants centralized processing by converting to district heating[4,5]. As a successful model, by means of energy conservation andexpansion of CHP (Combined Heat and Power production),Denmark has been able to maintain the same primary fuel

x: þ86 571 87951058.

consumption for a period of more than 30 years in spit e of about70% increase in GDP (Gross Domestic Product) [6].

A DH system mainly includes three parts: heating sources,heating consumers and heating pipe system [7]. Heating pipe sys-tem, which transports heat generated in heating sources to heatingconsumers is the most complicated part of a DH system.

Compared with civil DH systems, industrial DH systems haveseveral special characteristics: as the steam affects themanufacturing process and the production security of heatingconsumers, the heating pipe system needs to be more secure andreliable; ring-shaped heating networks with multiple heatingsources are commonly used in industrial DH systems to improvethe stability of heat supply; there might be heat-return steamconsumers, which produce steam and supply them to heatingnetworks in some working conditions; the high frequency and awide range of load variation of industrial heating consumers maylead to complicated steam flow regime, hydraulic maladjustmentor even steam stagnation in certain pipes, and steam stagnation

Nomenclature

V set of nodesE set of sectionsC set of fundamental circuitsM the number of sectionsN the number of nodesS the number of fundamental circuitsA matrix connecting nodes and sectionsB matrix connecting sections and circuitsq column vector of mass flow rate of nodesQ row vector of mass flow rate of sectionsq node mass flow rate (kg/s)Q section mass flow rate (kg/s)DP row vector of pressure drop in each sectionDPflow.m flow resistance pressure drop in Em (Pa)DPgra.m gravitational pressure drop in Em (Pa)DPdyn.m dynamic pressure drop in Em (Pa)DPfriction.m friction resistance in Em (Pa)DPlocal.m local resistance in Em (Pa)x total local resistance coefficientl0 friction resistance coefficientl length (m)r density (kg/m3)w velocity (m/s)D diameter (m)k RoughnessRe Reynolds numberPr Prandtl numberNu Nusselt numberF heat transfer rate (W/m)K heat transfer coefficient (W/m2 K)d thickness (m)h convection heat transfer coefficient (W/m2 K)l conductivity coefficient (W/m K)T temperature (K)t temperature (�C)

P pressure (MPa)ε emissivitys StefaneBoltzmann constant (W/m2 K4)y kinematic viscosity (m2/s)Nst number of steam trapsNhc number of heating consumersNhs number of heating sourcesI enthalpy (kJ/kg)DQ regulated flow rate (kg/s)DH loop closure (Pa)F area (m2)εtot given precision of overall calculationεst given precision of drainage calculationP

qnst theoretical total drainage of steam trapsNhc.re number of heat-return consumersNhc.use number of heating consumers which use steam

Subscripts and abbreviationsDH district heatingCIWH condensation-induced water hammern serial number of a nodem serial number of a sections serial number of a fundamental circuitp pipeisu insulationa airr radiationm.c condensation layer in Emm.ms main steam in Emsh superheated steamw saturated waternst serial number of a steam trapnhc serial number of a heating consumernhs serial number of a heating sourcei iteration number of flow regulationj iteration number of overall calculation

W. Zhong et al. / Energy 87 (2015) 566e577 567

would result in some serious accidents, such as CIWH (condensa-tion-induced water hammer).

For industrial DH systems, operators can hardly get the real-time steam flow regime in each pipe for the lack of relatedmeasuring devices to record the flow rate, temperature and pres-sure of steam. However, with the rapid development of computertechnology, the combination of the heating network hydrauliccalculation model, real-time steam parameters of heating sourcesand consumersmakes it possible to calculate the steam flow regimetheoretically online and realize operation optimization for heatingnetworks. In the online hydraulic calculations, more attentionshould be paid to the drainage of all steam traps, which is inducedby heat dissipation in pipes since it can reach 3e10% of the totalheat supply.

At present, when calculating the steam flow regime in the pipesof a ring-shaped network, graph theory [8e12] and Kirchhoff's law[9,13] are used to divide the network into several loops and toanalyze the hydraulic condition with numerical methods [14]respectively. Technical University of Denmark [15] and softwareTERMIS [16] put forward a “node method” to simulate the tem-perature dynamics of DH systems, Irina Gabrielaitiene et al. [17]validated this method with time dependent consumer data froma Naestved DH system. Stevanovicet et al. [18] presented a method

for numerical simulation and analysis of the steady state hydraulicsof complex pipeline networks, which was based on the loop modelof the network and the square roots method for solving the systemof linear equations. In terms of business software, FLOWRA32 cansimulate real-time data in heat exchange stations online andanalyze the energy consumption, TERMIS [16] can optimize theoperation of DH systems according to different working conditionsand ambient conditions. They both can guide the design andoperation of DH systems, save energy and reduce cost for heat-supply companies.

For the water hammer phenomenon in steam heating networks,M. H. Chaudhryet et al. [19] had described the features of waterhammer early in 1987. As one of the main reasons that leads towater hammer [20], CIWH is an undesirable side effect whichwould result in instability [21] and hydraulic shock [22,23], andCIWH may cause a serious damage to equipment and endangerlives of working staff [24]. Much of previous research on CIWHwasbased on RELAP5, but M. Valinciuset et al. [25] and W. Zhou et al.[26] both identified the limitations of RELAP5 in their researchprojects. Based on the study of R.W. Bjorge and P. Griffith [27], M. H.Chunet et al. [22] presented a computer code entitled “KAIST-CIWH” and the sample guide charts to find CIWH free regions for agiven combination of major flow parameters in a long horizontal

W. Zhong et al. / Energy 87 (2015) 566e577568

pipe considering the effect of heat transfer from the steam to thepipe wall, Kun He et al. [28,29] studied the water condensationrates in a single steam pipe theoretically and experimentally.

The previous research listed above mainly focused on the hy-draulic calculation of heating networks and CIWH phenomenon,and there has been limited research about online hydraulic calcu-lation of steam heating networks, the effect of drainage on theoverall calculation and operation optimization of ring-shapedsteam heating networks with multiple heating sources.

In this paper, a general algorithm to calculate the steam flowregime in pipes is presented, inwhich the drainage of steam traps istaken into consideration, and an operation optimization method isproposed to help eliminate steam stagnation through optimizingthe heat load distribution of each heating source. Based on theresearch aforementioned, a general software system entitled“HEATNET” has been developed and applied in Shanghai chemicalindustry park successfully.

Fig. 2. Directed graph G of the heating network of Fig. 1.

2. Calculation model of heating networks

2.1. Directed graph model

A heating network of any structure can be regarded as a networksystem, which is linked by “nodes” and “sections” according tocertain topological structure. Fig. 1 shows a simplified heatingnetwork, “nodes” represents points with mass flow rate in and out,such as heating sources, heating consumers and steam traps;“sections” represents pipes that connect two nodes together, inwhich there would be some additional resistance components withthe same flow rate of the pipe, such as elbows and valves. All thenodes and sections in a heating network can be described with setV and E respectively, the network system linked by V and E isdepicted as a “directed graph” G in Fig. 2. In a ring-shaped network,there may bemany circuits in the “directed graph”, inwhich severalcircuits are defined as “independent circuits” that contain all thesections in the main circuits of the directed graph with the mini-mum number of circuits. All the independent circuits in thedirected graph can be described with set C.

V ¼ fV1;…Vn;…VNgð1 � n � NÞ (1)

Fig. 1. Sketch map of a simplified heating network.

E ¼ fE1;…Em;…EMgð1 � m � MÞ (2)

C ¼ f C1;…Cs;…CSgð1 � s � SÞ (3)

Based on graph theory [8e12], the number of nodes N, thenumber of sections M and the number of independent circuits Smust satisfy:

S ¼ MeN þ 1 (4)

In Fig. 2, there are 36 sections, 35 nodes and 2 independentcircuits C1 and C2, different nodes and sections in Fig. 2 aredistinguished by different subscripts. C1 consists of E2, E3, E5, E6, E7,E9, E10, E12, E32, E33, E35 and E36, C2 consists of E7, E9, E10, E12, E13, E15,E16, E18, E20, E21, E23, E28, E29 and E31.

To describe the topological structure of a heating network, amatrix A with N rows and M columns is defined to describe therelationship between V and E, another matrix B with S rows and Mcolumns is defined to describe the relationship between E and C.The elements of both twomatrixes are defined in Eq. (5) and Eq. (7).

A ¼

266664a11 ::: a1 ::: a1M« « « « «

an1 ::: anm ::: anM« « « « «

aN1 ::: aNm ::: aNM

377775 (5)

anm ¼8<:

1; Vn is the beginning of section Em�1; Vn is the end of section Em0; Vn;Em

(6)

B ¼

266664b11 ::: b1m ::: b1M« « « « «

bs1 ::: bsm ::: bsM« « « « «

bS1 ::: bSm ::: bSM

377775 (7)

Fig. 3. Directed graph G.

W. Zhong et al. / Energy 87 (2015) 566e577 569

bsm ¼8<:

1; Em2Cs and in the same direction�1; Em2Cs but in different directions0; Em;Cs

(8)

Take directed graph in Fig. 3 as example, there are 6 nodes, 9sections and 4 independent circuits in the directed graph, matrix Aand B are shown in Eqs. (9) and (10).

A ¼

2666666664

E1 E2 E3 E4 E5 E6 E7 E8 E9V1 0 0 0 0 �1 0 0 �1 �1V2 0 0 0 0 1 1 1 0 0V3 �1 �1 0 0 0 0 0 1 0V4 1 1 0 0 0 �1 0 0 0V5 0 0 1 1 0 0 0 0 1V6 0 0 �1 �1 0 0 �1 0 0

3777777775

(9)

B ¼

266664

E1 E2 E3 E4 E5 E6 E7 E8 E9C1 1 �1 0 0 0 0 0 0 0C2 0 1 0 0 �1 1 0 1 0C3 0 0 1 0 1 0 �1 0 �1C4 0 0 �1 1 0 0 0 0 0

377775 (10)

Mass flow rate of nodes and sections are described in Eq. (11)and (12) respectively:

q ¼ ½q1;…qn;…qN�T (11)

Q ¼ ½Q1;…Qm;…QM� (12)

qn is defined in Eq. (13):

8<:

qn >0; mass flow rate inqn <0; mass flow rate outqn ¼ 0; no mass flow rate

(13)

For common heating consumers, qn > 0; for heating sources andsome special heating consumers called “heat-return” consumerswhich produce steam themselves and supply steam to the heatingnetwork, qn < 0, in Fig. 2, heating consumers V5 and V8 are heat-return consumers.

Fig. 4. Local structure of Fig. 1.

2.2. Hydraulic calculation of heating networks

From previous studies, fluid flowing in pipes must satisfy theKirchhoff's law [13]: flow rate conservation law of nodes and en-ergy conservation law of circuits.

The flow rate conservation law of nodes can be described as: thealgebraic sum of flow rate going into and out of any node is zero,which is written as:

AQ T þ q ¼ 0 (14)

The energy conservation law of circuits is: the algebraic sum ofall pressure drops around a closed path, or mesh, in the networkmust be zero:

BDPT ¼ 0 (15)

Where DP¼ [DP1, … DPm, … DPM] is the row vector to record thepressure drop in each section. DPm consists of three parts: flowresistance pressure drop DPflow.m, gravitational pressure dropDPgra.m and dynamic pressure drop DPdyn.m, in which DPflow.m con-sists of friction resistance DPfriction.m and local resistance DPlocal.m, asshown in Eq. (16) and (17):

DPm ¼ DPflow:m þ DPgra:m þ DPdyn:m (16)

DPflow:m ¼ DPfriction:m þ DPlocal:m (17)

In steam heating networks, as the steam density is compara-tively low and two-phase flow seldom takes place during normaloperation, DPgra.m and DPdyn.m can be ignored [13], the pressuredrop of Em can be described as:

DPm ¼�xm þ l0m

lmDm

�r2mw

2m

2(18)

Take a local structure in Fig.1 as example, as shown in Fig. 4, E5 isan expansion bend which is commonly used to balance the thrustinduced by thermal expansion and contraction of steam. There arefour 90� elbows in E5; the beginning node of E5 is the outlet of V4and the end node is steam trap V6. Local resistance x5 consists of theresistance of four elbows and the resistance of the V4. According tothe calculation standard recommended by former Soviet Union, x5is calculated as [30]:

x5 ¼ x5:elbow þ x5:tee ¼ 4� 0:25þ xh1þ

�pd0�þ 2ð1� pÞ2

i(19)

where:

p ¼ Q5

Q4(20)

d0 ¼�D4

D5

�2(21)

l05 ¼

8>>>><>>>>:

64Re5

; Re5 � 2320

1�1:14þ 2lg

D5

k

� ; Re5 >2320(22)

Where x is a coefficient related to d0, x5.elbow is the resistance of theelbow, x5.tee is the additional resistance induced by V4, and DP5would be calculated by Eqs. (18)e(22).

Fig. 6. Sketch map of heat dissipation in pipes.

W. Zhong et al. / Energy 87 (2015) 566e577570

2.3. Drainage calculation of steam traps considering heatdissipation

Traditional drainage calculation of steam traps is mainly basedon operation experience, and ambient conditions are not taken intoconsideration [29]. In this paper, heat dissipation from the steam inpipes to the ambient is considered, and a drainage calculationmethod of steam traps considering heat dissipation is presented;the drainage is dispersed to each moment as qn of steam traps andapplied to the total hydraulic calculation of the steam heatingnetwork, which is shown in Eq. (14).

Take a unit-length overhead steam pipe as example, the crosssection of the pipe is shown in Fig. 5. The heat transfer process fromsteam to the ambient includes:

1. Heat convection from flowing steam to the inner wall of thepipe;

2. Heat conduction in the pipe and insulation layer;3. Heat convection and radiation from the insulation layer to the

ambient [31].

The heat transfer rate of a section Em can be calculated as:

Fm ¼ K$p�Dm þ 2dp þ 2disu

�$ðTm � TaÞ (23)

K ¼ 1

1pDmhm

þ ln Dmþ2dpDm

2plpþ ln Dmþ2dpþ2disu

Dmþ2dp

2plisuþ 1

pðDmþ2dpþ2disuÞðhaþhrÞ

(24)

hm ¼ 0:023Re0:8m Pr0:3m lm

.Dm (25)

Rem ¼ wmDm=ym (26)

ha ¼ NualaDm þ 2dp þ 2disu

(27)

Rea ¼ wa�Dm þ 2dp þ 2disu

�na

(28)

Nua ¼��

0:43þ 0:5Re0:5a�Pr0:38a 1<Rea � 103

0:25Re0:6a Pr0:38a 103 � Rea <2� 105(29)

hr ¼ εs�T2isu þ T2a

�ðTisu þ TaÞ (30)

As shown in Fig. 6, Fm consists of two parts:

Fm ¼ Fm:c þ Fm:ms (31)

Fig. 5. Cross section of a steam pipe.

Fm.ms is the dissipation which leads to temperature drop of themain steam [32,33], andFm.cmakes near-surface steam in the pipescondense to water [29].

The condensation water in pipes would flow along with thesteam and drain out through a steam trap, then the drainage ofsteam traps can be calculated as:

qnst ¼P

Fm:clmIsh � Iw

� 10�3 ¼P

gFmlmIsh � Iw

� 10�3 ðn ¼ 1;2;3;…NstÞ

(32)In which g is the proportion coefficient, it is related to the

structure and working conditions of the heating network, it can beestimated from the analysis of operating data:

g ¼ f

XNhs

n¼1

qnhs;XNhc

n¼1

qnhc

!(33)

Take sections in Fig. 4 as example, the drainage of steam trap V6can be calculated as:

qnst ¼ gðF3l3 þ F4l4 þ F5l5ÞIsh � Iw

� 10�3 (34)

3. Model solution

3.1. Overall process of model solution

For a given heating network, a directed graph can be establishedbased on its specific structure. The aim ofmodel solution is to figureout the mass flow rate distribution of Q and q which can satisfy Eq.(14) and (15) simultaneously, and the methods of iterativeapproximation and distributed computing are used to solve theirimplicit nonlinear relationship.

For drainage calculation of steam traps, steam velocity and flowdirection in each pipe are required in Eqs. (23)e(32), and thedrainage calculation is divided into two steps: “rough calculationstep” and “precise calculation step”. The precise drainage calcula-tion is shown in Eq. (32), while the rough drainage calculation canbe described as:

qnst ¼PNhs

n¼1 qnhs �PNhc

n¼1 qnhcNst

ðn ¼ 1;2;3; :::NstÞ (35)

The overall process of model solution is shown in Fig. 7. Firstly,by Eq. (35) and given mass flow rate of heating sources and con-sumers, the initial flow rate distribution of nodes q(0) can be ob-tained as follow:

qð0Þ ¼hq1

ð0Þ;…qnð0Þ;…qNð0Þi

(36)

The initial flow rate distribution of sections Q(0) can be calcu-lated by the least-square algorithm [34] without defining the di-rection of steam in each section:

Fig. 7. Flow chart of the overall calculation.

W. Zhong et al. / Energy 87 (2015) 566e577 571

Q ð0Þ ¼hQ1

ð0Þ;…Qmð0Þ;…QM

ð0Þi

(37)

In this paper, the drainage of each steam trap qnst is the memberof the mass flow rate of nodes q, therefore, the change of drainagecalculation results would lead to the change of q. As shown in Fig. 7,the theoretical drainage of steam traps qnst ð1Þ calculated by Eq. (32)is more precise than qnst ð0Þ calculated by Eq. (35), then the massflow rate of nodes q and sections Q would be updated. The maxloop-closure algorithm is applied to regulate the mass flow ratedistribution of sections Q. Once the mass flow rate distribution ofsections Q(j) (j ¼ 1, 2, …) is confirmed through multiple iterations,comparisons between the drainage of each steam trap qnst ðjÞ andthe former one qnst ðj�1Þ will be made to find out if the error in thesetting range; if not, qnst ðjÞ would be recalculated to increase theprecision of the calculation, the iteration will continue until themass flow rate distribution of sections Q0 and nodes q' meet boththe precision requirements of εtot and εst, which means they canmeet the requirements of both Eqs. (14) and (15).

Q ′ ¼ Q10;…Qm

0;…Qm0 (38)

3.2. Flow regulation

The maximum loop closure of an independent circuit can bedescribed by Eqs. (39) and (40):

DQ ¼ � DHs:maxPMm¼1

�����bsmvDPmvQm

����� (39)

DHs ¼ bs1DP1 þ…bsmDPm þ…bsMDPM ¼ SbsmDPm (40)

DHs.max is the loop closure with maximum absolute value.Suppose:

km ¼�xm þ lmlm

Dm

��2rmF

2m (41)

Fm is the cross section area, Eq. (18) can be simplified as:

DPm ¼ kmQ2m (42)

Assuming that the steam density and the resistance coefficientsare constant, then km would be constant, and the regulated flowrate DQ can be written as:

DQ ðiÞ ¼ � DHði�1Þs:max

2PM

m¼1

�km���bsmQ ði�1Þ

m

���� (43)

Q ðiÞm ¼ Q ði�1Þ

m þ bsmdsDQ ði�1Þ (44)

W. Zhong et al. / Energy 87 (2015) 566e577572

If the updated flow rate distribution Q(i) modified by Eq. (43)

makes DHðiÞs:max <DHði�1Þ

s:max, make ds ¼ 1; otherwise, modify ds until

DHðiÞs:max <DHði�1Þ

s:max.

4. CIWH prevention

Compared to the single heating source or tree-shaped steamheating networks, ring-shaped steam heating networks withmultiple heating sources can improve the security and economy ofheating networks, while the complicated structure and changeableworking conditions would result in some serious accidents, such aswater hammer.

Water hammer may occur in every pipe that contains water andsteam [35]; especially, CIWH would occur frequently in steamheating networks. With the continuing operation of the heatingnetwork, the condensate induced by heat dissipation in pipeswould accumulate at the bottom of the pipe and form a layer whichis called the condensation layer. If steam stagnation takes placeunder certain working condition and the discharge capacity ofsteam traps is less than the quantity of condensate, the thickness ofthe condensation layer and the condensate rate would keepincreasing. When the condensation layer is thick enough, interfa-cial waves would be formed with increased steam velocities inthese steam stagnation pipes under another working condition[25], this phenomenon is called CIWH [36]. In this paper, aneffective method to prevent CIWH is presented by eliminatingsteam stagnation through optimizing the heat load of each heatingsource.

To meet a specific heat load requirement of heating consumers,the heat load distribution of heating sources is uncertain, whilethey must satisfy the equation below:

XNhs

n¼1

qnhs ¼XNhc

n¼1

qnhc þXNst

n¼1

qnst (45)

For different heat load distribution of heating sources, the flowrate of nodes would be different, and the flow rate distribution ofsections Q would change according to Eqs. (14) and (15), the steamvelocity wm in each section is:

wm ¼ 4Qm

.�prmD

2m

�(46)

Ifwm in certain pipe is lower than a setting valuewmin, as shownin Eq. (47), it implies that steam stagnation has taken place insection Em under this specific working condition.

wm <wmin (47)

If operators of heat supply companies regulate the flow ratedistribution in sections by changing the load of each heating sourceqnhs to make sure wm > wmin, CIWH could be prevented.

wmin is the velocity limit that can avoid steam stagnation, it isrelated to the anti-pressure property of tube material, structure ofthe heating network and steam parameters, which can be esti-mated from the analysis of operating data.

5. Software development

A general software system entitled “HEATNET” is developed byobject-oriented language Cþþ in this paper, which could be appliedto arbitrary structured heating networks based on the graphicalmodel method, the backend interface of HEATNET is shown in

Fig. 8, and the detailed section parameters dialog are shown inFig. 9.

HEATNET can read operating data of a heating network fromreal-time database and display key information, such as the pres-sure of heating consumers and steam velocities in pipes on thefront-end to help monitor the steam flow regime in heating net-works. Moreover, it can locate steam stagnation pipes quickly andaccurately with the establishedmodel and the real-time calculationresults, and then would trigger the alarm to operators and providethem with an effective method to eliminate the phenomenon. Theonline running process of HEATNET is shown in Fig. 10.

Before online hydraulic calculation, all structure parametersshould be input to HEATNET according to the actual structure of theheating network, such as the positions of each heating source,heating consumer and steam trap, the length and diameter of eachpipe, insulation condition of the pipes, etc. During online moni-toring, the data of mass flow rate of heating sources and heatingconsumers, pressure and temperature of heating sources and heat-return consumers should be read from the real-time database ofSCADA (Supervisory Control and Data Acquisition) system accord-ing to the measure point names.

6. Application and discussions

At present, HEATNET has been applied to the heating network inShanghai chemical industry zone of China. As one of the world'slargest industry parks, over 30 world-famous chemical companieshave settled in, including BASF SE (Badische Anilin-und-Soda-Fabrik), Bayer and Degussa, etc.

With data from real-time database of Shanghai chemical in-dustry zone and the calculation model introduced in this paper, thepracticability of HEATNET and the importance of drainage calcula-tion of steam traps are validated; factors that affect the drainage ofsteam traps are discussed; and the method of preventing CIWH byoptimizing the heat load of each heating source is proved to beeffective.

6.1. Error analysis of pressure and temperature calculation

In this paper, two working conditions shown in Table 1 arestudied.

For the two working conditions, both heating consumers BASFBACH (Badische Anilin-und-Soda-Fabrik Chemical Company) andIndustrial Gas are heat-return consumers. Under working condition1, the total heat supply of heating sources and heat-return con-sumers is:

XNhs

n¼1

qnhs þXNhc::re

n¼1

qnhc ¼ 110:4552t=h (48)

The total heat consumption is:

XNhc:use

n¼1

qnhc ¼ 101:7077t=h (49)

And the actual total drainage is:

XNhs

n¼1

qnhs þXNhc:re

n¼1

qnhc �XNhc:use

n¼1

qnhc ¼ 8:7475t=h (50)

Under working condition 2, the total heat supply, heat con-sumption and the actual total drainage are as follows:

Fig. 8. The backend interface of HEATNET.

Fig. 9. The input interface of heating source in Fig. 8.

W. Zhong et al. / Energy 87 (2015) 566e577 573

Fig. 10. Flow chart of the on-line calculation.

Table 1Working conditions 1 and 2.

Name Flow rate (t/h)

Condition 1 Condition 2

Source 1# 0 43.90721Source 2# 80.40004 29.15809Source 3# 0 69.22825HPS 12.86171 13.02745SBR 49.31501 49.82751BACH-Return 19.85505 16.41336Industrial Gas-Return 10.20011 14.51041Bayer 1# 18.01357 49.08110Bayer 2# 15.58498 47.68795SBPC 4.53050 3.52057SSMC 1.40190 1.61306…… …… ……

Table 2Pressure and temperature comparison of calculated and measured data.

Name Working condition Item

HPS 1 P (MPa)t (�C)

2 P (MPa)t (�C)

SBR 1 P (MPa)t (�C)

2 P (MPa)t (�C)

Bayer 1# 1 P (MPa)t (�C)

2 P (MPa)t (�C)

Bayer 2# 1 P (MPa)t (�C)

2 P (MPa)t (�C)

SBPC 1 P (MPa)t (�C)

2 P (MPa)t (�C)

SSMC 1 P (MPa)t (�C)

2 P (MPa)t (�C)

W. Zhong et al. / Energy 87 (2015) 566e577574

XNhs

n¼1

qnhs þXNhc::re

n¼1

qnhc ¼ 173:2173t=h (51)

XNhc:use

n¼1

qnhc ¼ 164:7576t=h (52)

XNhs

n¼1

qnhs þXNhc:re

n¼1

qnhc �XNhc:use

n¼1

qnhc ¼ 8:4597t=h (53)

Both working condition 1 and working condition 2 are taken asexamples to validate the practicability of HEATNET, comparisonsbetween calculated pressure and temperature by HEATNET andmeasured data from real-time database of both working conditionsfor part of heating consumers are shown in Table 2.

We can conclude in Table 2 that errors between the calculateddata and measured data are within a reasonable range, the de-viations could be caused by measured deviation and the pipeinsulation aging, but also it may induced in the process ofmodeling,since we focus on the main network flow status, the connectionpipes between the main network branch point and the heatingconsumers are simplified model, so it is feasible to use HEATNET tohelp heat-supply companies to judge if the steam parameterstransported to each heating consumer are in the proper range.

6.2. Drainage of steam traps

Factor analysis method is applied in this paper to analyze theeffect of air temperature ta and wind speed ua on the theoreticaltotal drainage

Pqnst, and the analysis contains two aspects:

(1) ua is constant and ta ranges from 0 �C to 40 �C:

Fig. 11 shows the variation of theoretical total drainageP

qnstalong with ta at the conditions of ua ¼ 2.7 m/s and ua ¼ 4.0 m/sunder working condition 1.

Measured data Theoretical data Ratio error

4.201 4.337 �3.24272.137 274.768 �0.97

4.237 4.271 0.80270.547 275.126 1.70

3.994 3.905 �2.23274.677 271.392 �1.20

4.091 4.175 2.05278.746 281.269 0.91

4.193 4.325 3.15253.578 264.020 4.12

4.256 4.142 �2.68267.351 260.247 �2.66

4.216 4.316 2.37270.204 267.406 1.04

4.315 4.336 4.87275.245 278.649 1.23

4.363 4.426 1.44255.824 263.811 3.12

4.361 4.234 �2.91259.324 254.125 �2.01

4.434 4.260 3.92271.819 276.495 �1.72

4.412 4.348 �1.45269.125 267.415 �0.64

Fig. 11.P

qnst e ta profile under working condition 1.

Fig. 12.P

qnst e ta profile under working conditions 1 and 2.

Fig. 13.P

qnst e ua profile under working condition 1.

Fig. 14.P

qnst e ua profile under working conditions 1 and 2.

W. Zhong et al. / Energy 87 (2015) 566e577 575

Fig. 12 shows the variation ofP

qnst along with ta at the condi-tion of ua ¼ 2.7 m/s under working conditions 1 and 2.

It can be concluded from Figs. 11 and 12 that when ua is con-stant,

Pqnst would decrease with the increasing of ta under the

same working condition; and the trend would be more obviouswhen the heat load is higher.

(2) Ta is constant and ua ranges from 1.6 m/s to 4.5 m/s:

Fig. 13 shows the variation ofP

qnst along with ua at the con-ditions of ta ¼ 0 �C and ta ¼ 25 �C under working condition 1.

Fig. 14 shows the variation ofP

qnst along with ua at the con-dition of ta ¼ 25 �C under working conditions 1 and 2.

It can be concluded from Figs. 13 and 14 that when ta is constant.Pqnst would increase with the increasing of ua under the same

working condition; and the trend would be more obvious when theheat load is higher.

From the analysis above, we know thatP

qnst has a closerelationship with the air temperature ta and the wind speed ua,and

Pqnst could even vary about 10% of the total heat supply when

ta and ua are different under the same working condition. There-fore, it is unreasonable to ignore or estimate the drainage of steamtraps without theoretical calculation considering ambientconditions.

Table 3Working conditions 3, 4 and 5.

Name Flow rate (t/h)

Condition 3 Condition 4 Condition 5

Source 1# 0 0 12.5Source 2# 87.37748 79.00180 74.87748Source 3# 0 0 0HPS 12.78729 12.68088 12.78729SBR 49.42872 49.14616 49.42872BACH-Return 20.13164 21.68473 20.13164Industrial Gas-Return 8.98247 9.11874 8.98247Bayer 1# 17.40832 16.23876 17.40832Bayer 2# 18.20041 12.18896 18.20041SBPC 6.98954 8.20395 6.98954SSMC 1.36139 1.40400 1.36139…… …… …… ……

Fig. 15. Local steam velocity (m/s) distribution under working conditions 3, 4 and 5.

W. Zhong et al. / Energy 87 (2015) 566e577576

6.3. Operation optimization of heating networks

Before the application of HEATNET in Shanghai chemical in-dustry zone, the heating network had once experienced a CIWHaccident.

From the database, we can read the working conditions beforeand after the CIWH accident, which are working condition 3 andworking condition 4 in Table 3, respectively. Fig. 15 presents thesteam velocity interface of the marked pipes with rectangle inFig. 8, which corresponds to working conditions 3, 4 and 5 inTable 3. In Fig. 15, under working condition 3 the minimum steamvelocities are only 0.007 m/s, which are lower than the settingvelocity limit wmin of this specific heating network. From theanalysis of historical data, the working condition of the wholeheating networks almost stays unchanged until working condition4 took place about one hour later, under which theminimum steamvelocities are 0.23m/s, sowe can affirm that steam stagnation is thereason of this CIWH accident.

To prevent CIWH, according to the method presented in thispaper, heat-supply companies should change the heat loads ofSource 1# and Source 2#, so that to switch working condition 3 toworking condition 5 in Fig. 15. With the heat load of each heatingconsumer remains unchanged, the minimum steam velocitycalculated by HEATNET of the whole heating network underworking condition 5 is 0.12 m/s, which is higher than wmin, thenCIWH could be prevented and the safety of the whole heatingnetwork would be guaranteed with the operation optimizationmethod.

7. Conclusions

In this paper, HEATNET has been developed and applied to theheating network in Shanghai chemical industry zone and con-nected to the real-time database to help manage the operation ofthe heating network. Conclusions from the practical applicationand analysis are summarized as follows:

(1) The drainage of steam traps should not be ignored since itcould reach a high ratio of the total heat supply, and for theclose relationship with the ambient conditions, it also shouldnot be estimated without theoretical calculation;

(2) The comparison between the calculation results of HEATNETand measured data from real-time database demonstratesthe reliability of the hydraulic calculation method;

(3) The method of eliminating steam stagnation through opti-mizing the heat load of each heating source proves to beoperable and efficient in preventing CIWH;

(4) The software systemHEATNET shows favorable commonalityand flexibility, which can help master the steam flow regimein pipes and optimize the production processes of heat-supply companies to achieve greater economic outcomes.

Acknowledgments

The work presented in this paper is supported by FundamentalResearch Funds for the Central Universities of China.

References

[1] Nielsen S, Moller B. Excess heat production of future net zero energy buildingswithin district heating areas in Denmark. Energy 2012;48:23e31.

[2] Jie PF, Tian Z, Yuan SS, Zhu N. Modeling the dynamic characteristics of adistrict heating network. Energy 2012;39(1):126e34.

[3] Rezaie B, Reddy BV, Rosen MA. An enviro-economic function for assessingenergy resources for district energy systems. Energy 2014;70:159e64.

[4] Lund H, M€oller B, Mathiesen BV, Dyrelund A. The role of district heating infuture renewable energy systems. Energy 2010;35(3):1381e90.

[5] Rezaie B, Rosen MA. District heating and cooling: review of technology andpotential enhancements. Appl Energy 2012;93:2e10.

[6] Lund H. Renewable energy strategies for sustainable development. Energy2007;32(6):912e9.

[7] Ajah AN, Mesbah A, Grievink J, Herder PM, Falcao PW, Wennekes S. On therobustness, effectiveness and reliability of chemical and mechanical heatpumps for low-temperature heat source district heating: a comparativesimulation-based analysis and evaluation. Energy 2008;33(6):908e29.

[8] Brkic D. An improvement of Hardy Cross method applied on looped spatialnatural gas distribution networks. Appl Energy 2009;86(7e8):1290e300.

[9] Boulos P, Altman T. A graph-theoretic approach to explicit nonlinear pipenetwork optimization. Appl Math Model 1991;15(9):459e66.

[10] Han ZY, Weng WG. An integrated quantitative risk analysis method for nat-ural gas pipeline network. J Loss Prev Process Ind 2010;23(3):428e36.

[11] Epp R, Fowler AG. Efficient code for steady-state flows in networks. J HydraulDiv 1970;96(1):43e56.

[12] Kesavan H, Chandrashekar M. Graph-theoretical models for pipe networkanalysis. J Hydraul Div 1972;98(2):345e64.

[13] Yan AB, Zhao J, An QS, Zhao YL, Li HL, Huang YJ. Hydraulic performance of anew district heating systems with distributed variable speed pumps. ApplEnergy 2013;112:876e85.

W. Zhong et al. / Energy 87 (2015) 566e577 577

[14] Cross H. Analysis of flow in networks of conduits or conductors. Eng Exp Stn1936;286:3e29.

[15] Benonysson A. Dynamic modelling and operational optimization of districtheating systems. PhD thesis. Denmark: Technical University of Den-mark;1991.

[16] TERMIS. Manual version 2.0, 7-Technologies A/S. Available from:. March 2003.http://www.7T.dk.

[17] Gabrielaitiene I, Bøhm B, Sunden B. Modelling temperature dynamics of adistrict heating system in Naestved, Denmarkda case study. Energy ConversManag 2007;48(1):78e86.

[18] Stevanovic VD, Prica S, Maslovaric B, Zivkovic B, Nikodijevic S. Efficient nu-merical method for district heating system hydraulics. Energy Convers Manag2007;48(5):1536e43.

[19] Chaudhry MH. Applied hydraulic transients. 2nd ed. New York, USA: VanNostrand Reinhold Co; 1987. p. 35e9.

[20] Yow W, Duyne DAV, Chiu C. Analysis of root causes of water hammer. In: Thethird international topical meeting on nuclear power plant Thermal-Hydraulics and operations (NUPTHO-3); 1988. p. 103e9.

[21] Ashraful Alam MM, Matsuo S, Setoguchi T. Effect of non-equilibrium homo-geneous condensation on the self-induced flow oscillation of supersonicimpinging jets. Int J Therm Sci 2010;49(11):2078e92.

[22] Chun MH, Yu SO. A parametric study and a guide chart to avoid condensation-induced water hammer in a horizontal pipe. Nucl Eng Des 2000;201(2e3):239e57.

[23] Kirsner W. Waterhammer. HPAC heating, piping and air conditioning. 1999.p. 113e22.

[24] Milivojevic S, Stevanovic VD, Maslovaric B. Condensation induced waterhammer: numerical prediction. J Fluids Struct 2014;50:416e36.

[25] Valincius M, Seporaitis M, Kaliatka A, Pabarcius R, Gasiunas S, Laurinavicius D.The concept and RELAP5 model of thermal-hydraulic system, employing arapid condensation for coolant circulation. Heat Transf Eng 2014;35(4):327e35.

[26] Zhou W, Wolf B, Revankar S. Assessment of RELAP5/MOD3.3 condensationmodels for the tube bundle condensation in the PCCS of ESBWR. Nucl Eng Des2013;264:111e8.

[27] Bjorge RW, Griffith P. Initiation of waterhammer in horizontal and nearlyhorizontal pipes containing steam and subcooled water. J Heat Transf1984;106(4):835e40.

[28] Yan X, Li J, He K, Feng ZP, Fan YZ, Shang XY. Investigations on watercondensation in the main steam pipe of a pressurized water reactor nuclearpower station: part I e experimental tests. Proc Inst Mech Eng Part A J PowerEnergy 2014;228(4):380e90.

[29] He K, Li J, Yan X, Feng ZP, Fan YZ, Shang XY. Investigations on watercondensation in the main steam pipe of a pressurized water reactor nuclearpower station: part II e theoretical studies. Proc Inst Mech Eng Part A J PowerEnergy 2014;228(4):391e401.

[30] Code for design of thermal power plant steam/water piping. Ministry of po-wer industry of the people's Republic of China. DL/T 5054e1996.

[31] Holman JP. Heat transfer. 6th ed. New York, USA: McGraw-Hill Book Com-pany; 1986.

[32] Colburn AP, Hougen OA. Design of cooler condensers for mixtures of vaporswith noncondensing gases. Ind Eng Chem 1934;26:1178e82.

[33] Gerber AG, Kermani MJ. A pressure based EulerianeEulerian multi-phasemodel for non-equilibrium condensation in transonic steam flow. Int J HeatMass Transf 2004;47(10):2217e31.

[34] Al-Omari AS, Chaudhry MH. Unsteady-state inverse chlorine modeling innetworks. J Hydraul Eng 2001;127(8):669e77.

[35] Dirndorfer S, Doerfler M, Kulisch H, Malcherek A. Condensation induced waterhammereoverview and own experiments. Kerntechnik 2012;77(2):122e7.

[36] Griffith P. Screening reactor steam/water piping systems for water hammer.Nuclear regulatory commission. Washington, DC (United States): Div. ofSystems Technology; 1997.