Integrated Design of Smart Train Scheduling, Use of ...

IEEJ Journal of Industry ApplicationsVol.8 No.6 pp.893–903 DOI: 10.1541/ieejjia.8.893

Paper

Integrated Design of Smart Train Scheduling, Use of Onboard EnergyStorage, and Traction Power Management for Energy-Saving Urban

Railway Operation

Warayut Kampeerawat∗a)Student Member, Takafumi Koseki∗ Member

(Manuscript received Nov. 11, 2018, revised May 9, 2019)

This paper presents an integrated design of train scheduling, use of onboard energy storage, and traction power man-agement for urban railways. The proposed design aims to integrate the design of train operation and infrastructure toimprove energy-saving operation and the flexibility of energy management. The design problem is formulated as theminimization of the energy supplied from substations and the energy capacity of onboard energy storage. By varyingthe weighting factor, energy-saving purpose and cost-saving purpose can be compromised. To demonstrate the per-formance of the proposed design, numerical case studies are performed and evaluated on the Bangkok Mass TransitSystem. From the comparisons of nominal operation and design operating conditions, it is seen that the energy-savingperformance is improved by up to 9.65% and the peak power at a substation is reduced by approximately 40%. Thedesign scenario can be simply classified into cheap, moderate, and expensive designs depending on the appropriateadjusting weighting factor. Furthermore, the effect of pantograph voltage is evaluated and discussed. From the results,it is seen that the energy-saving performance is reduced by approximately 1% due to the fluctuation of the pantographvoltage. Even though the variation of pantograph voltage affects the design scheduling, a small deviation in the runningtime in some sections must be allowed. The proposed design still provides considerable improvement with regard tothe energy-saving operations. The proposed design employs an offline design and planning because the design processrequires considerable computation time.

Keywords: onboard energy storage, railway power management, regenerative power, train scheduling

1. Introduction

Nowadays, railway systems are one of the most efficientforms of transportation providing an effective use of energyand are a good solution to traffic problems. Railway sys-tems can still utilize energy more effectively as the bulk ofenergy from various sources is required for operation. There-fore, the proposals on energy management and energy-savingoperations are interesting issues. Design of driving strategy,train scheduling, vehicle and relevant systems are proposedas ways to improve energy usage and management in rail-ways (1) (2).

In modern railways, regenerative energy management hasbecome a key for energy-saving operations. Due to theadvancement of power electronic technology, regenerativebraking systems can recover considerable traction energy asregenerative energy. Basically, regenerative energy is used bythe train itself or it can be effectively managed by interchang-ing between trains based on smart train scheduling, storingand recycling based on energy storage system, and being fedback to the utility grid through an inverting substation (1) (3).

When considering single train operation, driving strategy

a) Correspondence to: Warayut Kampeerawat. E-mail: [email protected]∗ Department of Electrical Engineering and Information Systems,

Graduate School of Engineering, The University of Tokyo7-3-1, Hongo, Bunkyo-ku, Tokyo 113-8656, Japan

design provides the reduction of traction energy consumedby each train, but the mutual operation among multiple trainstends to be quite complicated when included in the de-sign (4)–(6). Therefore, regenerative power management has notbeen considered in such proposals. To improve the utiliza-tion of regenerative energy including mutual operating con-ditions of multiple vehicles, novel proposed strategies deal-ing with multi-train operation have been proposed in (7)–(11)based on maximizing the possibility of exchanging regenera-tive power among trains.

Furthermore, to enhance the flexibility of energy manage-ment, additional systems are introduced to develop existingrailways and designing future railways. An energy storagesystem (ESS) is an effective option for energy management.Onboard energy storage aims to increase flexibility of energymanagement, reducing peak power, and enabling operationin non-electrified sections (12)–(14). A numerical study regard-ing the installation of onboard batteries on trains operatedin Bangkok Mass Transit System (BTS) showed the possi-bility for saving energy costs (15). Wayside energy storage isexpected to not only support energy-saving purpose but alsoenabling voltage stabilization. Optimizing control strategy,capacity and location of ESS have been proposed in (16)–(19). Energy storage provides effective energy management.Considerable additional cost may be considered for practicaldesign. Installing an inverting substation allows for feedingof surplus regenerative energy back to the utility grid. The

c© 2019 The Institute of Electrical Engineers of Japan. 893

Integrated Design of Smart Train Scheduling, Use of Onboard Energy（Warayut Kampeerawat et al.）

operator of the connected grid must ensure that the grid canstill work properly. The coordination between railway sys-tems and the utility grid must be studied before practical ap-plication (20) (21).

Combining multiple strategies into the one design pro-cess is supposed to provide better improvements of energy-saving operations and more efficient energy management (1).A combination of multiple methods may be carried out inan non-integrated way, i.e. each process is performed inde-pendently, or in an integrated way, i.e. multiple processes areperformed simultaneously. A comparison of non-integratedand integrated design has been performed in (22) based onintegrating train scheduling and installing onboard energystorage. Integrated design provides better energy-saving per-formance when compared with the non-integrated method.An integrated design of energy-saving driving pattern andtrain scheduling based on energy-efficient algorithm was in-troduced in (23). The proposed design provides fast calcu-lation but neglected exchanging regenerative power amongtrains. To consider effective energy management in multi-train operation, a cooperative train control model to designenergy-saving train scheduling based on simple search algo-rithm was developed by (24). Due to the complexity of in-tegrating various factors and parameters into the same prob-lem, some metaheuristic methods, e.g. Genetic Algorithm,are employed to solve the problem (25) (26). A two-layer opti-mization including timetable and driving strategy was pre-sented in (25). The running times of each train were adjustedto minimize energy consumption based on the idea of syn-chronizing power-time profiles by using a simple estimationof energy. Moreover, an integrated optimization of drivingpattern and timetable was proposed in (26).

To increase the flexibility of power management, inte-grated design of train operation and use of additional systemswere proposed by (8), (27). An optimal design of speed pro-files with consideration of regenerative energy recovery wasproposed in (27). To manage regenerative braking energy, theapplication of onboard ESS was considered in the design ofenergy-saving speed profiles. Design of train scheduling andwayside ESS with minimizing energy supplied from substa-tions was proposed in (28) (29). The timetable parameters,location and capacity of ESS was optimized based on Ge-netic algorithm.

This paper presents an integrated design method for effi-cient railway operation by combining design of train schedul-ing, use of onboard energy storage, and traction power man-agement. The proposed design aims to improve energy-saving operation by smart train scheduling based on maxi-mizing regenerative energy usage among trains. In addition,enhancing the flexibility of energy management is enabledby active use of onboard ESS. Therefore, the design of trainoperating condition and design of infrastructure will be si-multaneously considered in the same process. The optimiza-tion problem is formulated based on reducing energy sup-plied and minimizing cost of ESS. Combining train schedul-ing and ESS makes the problem more complicated to solvedue to the complexity of objective function, the large num-ber of decision parameters, and constraints. Therefore, Ge-netic algorithm is developed to solve the problem formulatedin the proposed design. This paper is organized as follows.

The integrated design for determining the operating scenar-ios and the details of problem formulation will be explainedin the second part. The main objective of the design is tominimize the energy supplied and the capacity of energystorage. Determining appropriate weighting factors, designresults can be obtained with arbitrary compromise betweenenergy-saving objectives and cost-saving objectives. In thethird section, numerical case studies based on Bangkok RapidTransit System (BTS) have been performed to verify the pos-sibility of applying the proposed method in a practical sys-tem. After showing all evaluating results, the discussion onthe results and concerns about the effect of pantograph volt-age have been evaluated and compared with an ideal case inthe fourth section. Finally, the paper will be concluded, andfuture works will also be mentioned.

2. Proposed Integrated Design

2.1 Integrated Design Concept The proposed inte-grated design aims to design train scheduling, onboard ESS,and strategy of traction power management by simultane-ously optimizing timetable parameters and the capacity ofESS. Employing onboard ESS, the proposed management oftractive power and regenerative power can be explained byFig. 1. In braking mode, regenerative power will be used byonboard auxiliary systems, stored in onboard ESS, then thesurplus regenerative power will be sent to nearby trains viacatenary. If such regenerative power cannot be absorbed bythe catenary, it will be wasted in the resister.

Besides increasing the use of regenerative power by thetrain itself and nearby trains, another purpose of the pro-posed design is to reduce the requirements of high capacityof onboard ESS. In powering mode, trains mainly consumepower from the catenary and additional power from onboardESS. For the proposed integrated design, total energy sup-plied from power substations is to be reduced by means ofmaximizing regenerative power usage. In the design of trainscheduling, adjusting timetable parameters, e.g. running timeand dwell time, are performed to increase the possibility ofexchanging power among trains. While the timetable is beingdesigned, the appropriate capacity of ESS is simultaneouslydetermined based on energy-saving and cost-saving objec-tives.

Regarding the mutual relationship between train schedul-ing and ESS capacity, the strategy for utilizing regenerativepower is based on onboard ESS first, then the remaining

Fig. 1. Power Management Scheme with Onboard ESS

894 IEEJ Journal IA, Vol.8, No.6, 2019


(a) Integrated Design (b) Non-integrated Design

Fig. 2. Integrated Design vs Non-integrated Design

regenerative power will be managed by train scheduling.Therefore, the capacity of ESS affects the remaining regener-ative power leading to the change of train scheduling. WhenESS’s capacity is large, the remaining regenerative powerwhich can be exchanged among trains is supposed to besmall. Therefore, ESS’s capacity is to be minimized to pro-vide the possibility of obtaining effective scheduling and sav-ing the ESS’s cost

The advantage of integrated design has been mentionedin (22). The non-integrated design is formulated as a sim-plified optimization and compared with the Integrated de-sign. Basically, the non-integrated approach aims to opti-mize timetable parameters and capacity of ESS sequentially.The basic concept of Integrated design and Non-integrateddesign can be illustrated as shown in Fig. 2. The numericalcase studies showed that the integrated design provides bet-ter energy-saving performance than the non-integrated designdoes.2.2 Problem Formulation The proposed integrated

design is formulated as an optimization problem having ob-jective function as shown in equation (1). The main objectiveaims to minimize energy supplied from traction substationsand energy capacity of ESS with a variable weighting factor(w).

min f (Tr,Td,Ness) = wEsub

Esub,base+ (1 − w)

Eess

Eess,max

· · · · · · · · · · · · · · · · · · · · (1)

The constraints for optimization problem are determined asfollows.

Headway limit: Th,min ≤ Th ≤ Th,max

Dwell time limit: Tda,min ≤ Tda ≤ Tda,max

Running time limit: Tr,a→b,min ≤ Tr,a→b ≤ Tr,a→b,max

Trip time: Ttrip,min ≤ Ttrip ≤ Ttrip,max

Regenerative limit: Ttr,reg ≤ Vreg,max

ESS charge and discharge: SOCmin ≤ SOC ≤ SOCmax

Where

Td = [Td1, Td2, . . . , Tdn] : Dwell time,

Tr = [Tr,1→2, Tr,2→3, . . . , Tr,(n−1)→n] : Running time

Th: Headway(s), Tda: Dwell time at passenger station a(s),Tr,a→b: Running time from passenger station a to stationb(s), Ttrip: Trip time for single journey of a train (s), Esub:Estimated total energy supplied from substations (kWh),Esub,base: Estimated total energy supplied from substations(kWh) in case of nominal operating condition, Ebrake: Es-timated total energy generated from electrical brake system(kWh), Ereg: Estimated total regenerative energy utilized by

trains (kWh), Eess: Total energy capacity of energy storagesystem (kWh), Eess,max: Maximum energy capacity of en-ergy storage system (kWh) which can be installed in the sys-tem, Ness: Number of energy storage modules, SOC: Stateof charge of ESS, Vtr,reg: Voltage of train at pantograph inregenerative mode, Vreg,max: Maximum regenerative voltage,w: weighting factor.

The first term in the objective function represents theenergy-saving objective and the second one represents thecost-saving objective. The weighting factor can be variedfrom 0 to 1 and it is to determine how much each objective isconcerned. (i.e., high w means high concern on energy sav-ing entailed expensive design conditions, while low w meansless concern on energy-saving purpose leading to cheap de-sign conditions). The design parameters consist of runningtime, dwell time, and energy capacity of energy storage.2.3 Solving Algorithm Dealing with optimizing

timetable parameters and ESS based on evaluating energy viamultiple train operations, the proposed optimization problemrequires efficient solving methods to cope with the objectivefunction and constraints. Genetic Algorithm (GA) is selectedto solve the problem due to its performance on finding thesolution of non-convex problems including the complicatedconstraints and the flexibility of expanding variable sizes (26).The basic algorithm of GA is shown in Fig. 3(a). To solvethe proposed design problem, the chromosomes are definedas running times, dwell times and energy capacity of ESS.Calculating fitness function shown in Fig. 3(b) is based onthe result of power flow calculation of each time step over aspecific period. The terminating criteria is the stall genera-tion satisfying the tolerance of fitness function allowing thealgorithm to stop when the average relative change in the fit-ness function exceeds the predetermined number of the stallgeneration limit.2.4 Estimation of Energy The estimation of relevant

energy is based on the following assumptions.- For train movement calculation, motor efficiency is as-

sumed as constant.- Tractive and brake effort’s curve, gradient, curvature, and

speed limit are included in the calculation.- For power flow calculation, resistance per length of run-

ning rail and catenary are assumed as constant at a specifictemperature.

- For calculating fitness function, the effect of pantographvoltage on tractive performance is neglected. Therefore, trainmovement calculation and power flow calculation can be per-formed separately.

Based on these assumptions, the process of calculatingfitness function can be simplified by using a precalculateddatabase as shown in Fig. 3(b). To estimate energy, the powerprofile versus time of each train and power supply from sub-stations will be first calculated, then the relevant energy willbe estimated by numerical integration. For generating thepower profile of each train, train movement calculation isperformed by neglecting the effect of voltage to train per-formance. By assuming that all trains have the same powerprofile, the power profiles of multiple train operations can besimply generated by shifting the time coordinate based oncorresponding timetable parameters.

To evaluate energy supplied from the substation, the power



(a) Genetic algorithm process

(b) Calculation of fitness function

Fig. 3. Solving algorithm based on GA

flow calculation is implemented based on algorithm proposedby (30) (31). To determine nodal voltage and current at anypoint in the system, the power flow calculation considersexchanging regenerative power among trains and power re-lated to ESS’s operation. Accordingly, the power flow resultsare used for estimating regenerative energy, charged and dis-charged energy of ESS, and energy supplied by substations.

To solve for the power flow calculation, the electrical mod-els of system components are explained as follows. Thetrain’s power and current are calculated by equation (2) and(3), respectively.

Ptr =

⎧⎪⎪⎪⎪⎨⎪⎪⎪⎪⎩

FT×vη+ Paux; Powering mode

Paux; Coasting modeη × FB × v + Paux; Braking mode

· · · · · · · · · · · · · · · · · · · · (2)

Itr =Ptr

Vpant − Vrail· · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · (3)

Where V pant: Nodal voltage at pantograph, Vrail: Nodalvoltage at rail conductor, FT : Tractive effort, FB: Brake ef-fort, Paux: Auxiliary power, Itr: Train’s current, Ptr: Train’spower, η: Motor’s efficiency.

Basically, the power substation is considered as non-inverting substation which cannot absorb regenerative powerfrom the braking train. When, the negative current is de-tected, the substation will be modeled as a large resistanceto limit the current.

Fig. 4. The flowchart of control algorithm for onboardESS

Fig. 5. An example of total substation power’s profile vstime

To evaluate energy, power and SOC of ESS, the amount ofpower in charging mode or discharging mode for each oper-ated train is estimated based on the control strategy shown inFig. 4.

In powering mode, discharging power (Pess(t)) and currentenergy stored (Eess(t)) are estimated based on total powerconsumed by the train (Ptr(t)) and the SOC of ESS. ESS willsupport the power consumed by the train with the maximumdischarging capacity (Pdis,max). In braking mode, discharg-ing power (Pess(t)) and current energy stored (Eess(t)) are es-timated based on total regenerative power of the train (Preg(t))and the SOC of ESS. ESS will charge the regenerative powerwith the maximum charging capacity (Pchg,max).2.5 Evaluation of Regenerative usage and Energy-

saving Performance1) Time period for evaluating energy quantityRelevant energy will be calculated by integrating a power

profile in a 1-hour period. Therefore, the energy quantitywill be presented in kWh/hr. To demonstrate the evaluatingperiod for calculating energy, the example of the power pro-file shown in Fig. 5 is illustrated with 60-minute evaluatingperiods.

2) Utilization of regenerative energy is defined as the ratioof energy recovered from the brake operation and total brakeenergy

%Ereg =Ereg

Ebrake× 100 · · · · · · · · · · · · · · · · · · · · · · · · · · · (4)

3) Energy-saving performance is defined as the percentageof substation energy which can be decreased compared withsubstation energy of nominal operation.



%Esave =(Esub,base − Esub,case i)

Esub,base× 100 · · · · · · · · · · · · (5)

2.6 The effect of Pantograph Voltage to Train Perfor-mance Basically, variation of pantograph voltage affectsthe performance of traction motor by changing the tractiveeffort characteristics which demonstrate the relation of speedand traction force (32). The relationship of traction force char-acteristics and pantograph voltage are described based on theequations (6) and (7).

vl = vlow · VVN· · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · (6)

vh = vhigh · VVN· · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · (7)

Where, vl, vh, vlow and vhigh are referred to the tractive ef-fort characteristics of traction motor. vl = low corner speedat current pantograph voltage, vh = high corner speed at cur-rent pantograph voltage, vlow = low corner speed at nominalpantograph voltage, vhigh = high corner speed at nominal pan-tograph voltage, V = pantograph voltage at current operatingcondition, VN = nominal pantograph voltage.

3. Numerical Case Studies

3.1 Case Study Information Bangkok Rapid Tran-sit System (BTS-Silom line), an elevated urban electric rail-way operated in Bangkok, Thailand, was selected for demon-strating the integrated design case. There are 13 passengerstations along a 13-km-long double track and seven tractionsubstations shown in Fig. 6. The system parameters for cal-culation are shown in Table 1.

3.2 Nominal Operating Condition For comparisonof the designed results, the nominal operation is assumed asthe operation of 5% time reserve mode or 1.05 of minimumrunning time. The nominal operating conditions and evalua-tion of relevant energy are shown in Table 2 and Table 3. The

Fig. 6. BTS silom line’s route map

Table 1. Basic information of BTS silom line

Table 2. Estimated energy for nominal operation without onboard energy storage

Table 3. Timetable parameters of nominal operation and Maximum/Minimum boundary (all values are in seconds)



(a) Peak hour operation (b) Off-peak hour operation

Fig. 7. Operating timetable for nominal operating condition

(a) Peak hour operation (b) Off-peak hour operation

Fig. 8. Nominal speed profile, speed limit, and gradient data VS Position

operating diagram (OD) are shown in Fig. 7. The nominalspeed profiles are shown in Fig. 8.3.3 Preparing Database of Speed and Power ProfileBased on the timetable parameters mentioned in Table 3,

the profile for all possible running time in the specified rangecan be calculated and kept in the profile database for furtheruse in the relevant calculating process. Because the weightof onboard ESS increases the total weight of the vehicle, theprofiles of trains with different capacities of ESS will be cal-culated as different cases. The scenarios for calculating pro-file are considered as follows.

- All possible running times in a feasible boundary (onlyin decimal value) in all operating sections along the route (12sections)

- Additional weights due to different capacities of ESS (1-10 kWh, decimal value)

- Different passenger loads due to traffic conditions (2 con-ditions, peak and off-peak conditions) and two running direc-tions (2 directions, southbound and westbound)

The example of speed profile and power profile in the pre-calculated database are shown in Fig. 9.3.4 Numerical Results The case studies are simu-

lated with two different traffic conditions and nine differentweighting factors as shown in Table 4. There are 18 casesin total. Case no.1–9 are performed in peak hour period andcase no.10–18 are in off-peak hour period. The specifica-tion for onboard ESS is an electrical double layer capacitor(ELDC) having 1 kWh, 300 kW, 428 kg per module (33). Theminimum SOC and maximum SOC of ESS are set at 25%

Fig. 9. Speed and Power profile in peak hour, sectionno.1 (W01→ CEN), 1 module of onboard ESS

and 100%, respectively. The maximum regenerative voltageis limited at 900 volts and the minimum pantograph voltage is500 volts. Parameters of GA are as follows: Maximum num-ber of generations is 200, population is 100 times the numberof variables, mutation rate is 0.8, crossover rate is 0.2. Stallgeneration limit is 30 with 1e-6 of fitness function tolerance.The numerical case studies are performed by using MATLABrun on a quadcore, intel core i7 with 16 GB RAM.

Variation of weighting factor from low values to high val-ues demonstrates different design scenarios ranging fromcheap infrastructure design to expensive infrastructure de-sign. The results obtained from the proposed design, %Esave



Table 4. Designed results and evaluation of %Esave and%Ereg of all case studies

and %Ereg are shown in Table 4. The design parameters forscheduling (i.e., OD1-OD7) are shown in Table 5.

The proposed method assumes the designed condition ex-cluded the effect of pantograph voltage to simplify the solv-ing process of the optimization problem, but, in practicaloperations, variation of pantograph voltage affects the per-formance of the train. Therefore, in Table 4, %Esave and%Ereg are also evaluated by including voltage’s effect (as-sumed to be the same as in a practical operation) for com-parison. The calculation is based on the same design drivingstrategy, ESS’s capacity and ESS’s control strategy.

For peak hour conditions, when the weighting factor is var-ied from 0.1 to 0.9, only four different scheduling results(OD1, OD2, OD3 and OD4) are obtained. From the resultsshown in Table 4, 4.12% of supplied energy can be savedwith the cheapest design, OD1 with 1 kWh of ESS, and upto 5.87% of energy can be saved with the most expensive de-sign, i.e., OD4 with 2 kWh of ESS.

For off-peak hour conditions, only three different schedul-ing (OD5, OD6, and OD7) are obtained from varying weight-ing factor. From the results, 5.70% of supplied energy can besaved with the cheapest design, i.e. OD5 with one ESS, andup to 9.65% of energy can be saved with the most expensivedesign, i.e., OD7 with 3 kWh of ESS.

The variation of the weighting factor affects the perfor-mance of designed results as shown in Fig. 10. The plot ap-pears to reflect the low sensitivity of %Esave and %Ereg to thechange of the weighting factor, because the same designedresults are obtained from multiple weighting factors. Thereason for low sensitivity is the resolution of ESS capacity

which is determined as 1 kWh to reflect the practical specifi-cation of ESS modules used in practical railway application.

The sensitivity plot of %Esave and %Ereg to the change ofweighting factor demonstrates the tendency of designed per-formance. When the weighting factor is changed by a consid-erable quantity, the different design results are obtained withbetter performance. When the weighting factor is changedfrom the minimum value to the maximum value, the weight-ing factor obviously affects the performance of the design re-sults. Based on the weighting factor, the design scenarios canbe classified as cheap, moderate, and expensive infrastructuredesign.

When the effect of pantograph voltage is considered inthe calculation, %Esave is reduced by 0.57% in peak hourcases and 0.93% in off-peak hour cases. The degradationof the designed performance results from the small devia-tion in running time which affects the efficiency of designedscheduling. The speed profile, pantograph voltage profile,the train’s power, and SOC for designed operating conditionswith and without the effect of pantograph voltage are com-pared in Figs. 11 and 12. The voltage profile demonstratedin Figs. 11(a) and 12(a) shows the variation of pantographvoltage of a train with corresponding speed profile. The pan-tograph voltage always satisfies the regenerative voltage limitand minimum voltage.

The speed profile and voltage profile for the design oper-ating condition of OD4 (peak hour) and OD7 (off-peak hour)with and without the effect of pantograph voltage are com-pared in Figs. 11(a) and 12(a), respectively. When the de-sign scheduling is applied with inclusion of the pantographvoltage, the deviation of running time in some sections areobserved. Such deviation entails the error in design schedul-ing and the decrease of %Esave. The traction power and SOCof ESS with the effect of pantograph voltage are showed inFigs. 11(b) and 12(b).

From the comparisons shown in Figs. 13 (a) and (b), theeffect of pantograph voltage degrades both the %Esave and%Ereg. The effect of voltage is more obvious when the ca-pacity of ESS is less. Therefore, the performance in a cheapdesign condition entails higher sensitivity to pantograph volt-age than in expensive design conditions.

When considering the power supplied by each power sub-station, the designed operating conditions provide a consid-erable reduction in peak power at substations, especially inpeak hour when the peak power. As shown in Fig. 14, peakpower can be reduced by approximately 43% in peak hour atsubstation no.2 and by approximately 41% in off-peak hourat substation no.3.

Basically, the onboard ESS can effectively suppress peakpower at a substation. When the scheduling is changed, theoverlap of peak power at each substation is also changed. Theproposed design is aimed to reduce the capacity of ESS, butthe suppression of peak power has not been included in theobjective function yet. The increase of peak power comparedwith nominal operation may be obtained in some cases be-cause the capacity of ESS may not be large enough to supportpeak power at some substations in some operating periods.Therefore, it is possible that peak power of the designed casecan be higher than that of a nominal case in some substations.

Although, in case of OD1, the peak power at substation 6 is



Table 5. Designed timetable for all case studies (all values are in seconds)

(a) %Esave with varying weighting factor (b) %Ereg with varying weighting factor

Fig. 10. %Esave and %Ereg vs Weighting Factor without effect of pantograph voltage

increased less than 0.5 MW when compared with the nominalcase, the peak power at substation 2 and 3 decreased consid-erably. In the case of off-peak hour, the peak power at substa-tion 5 increased in both cases, OD5 and OD7, while the peakpower at substation 3 was suppressed considerably.

4. Discussions

4.1 Determining Weighting Factor The weightingfactor is designed for compromising the energy-saving andthe cost-saving objectives. The variation of weighting fac-tor provides different design conditions depending on the re-quirements of the operator. The large weighting factor rep-resents the expensive infrastructure design, while decreasing

the weighting factor tends to reduce the cost of infrastructureor additional systems. However, various designed results ob-tained from varying weighting factors may overlap depend-ing on the resolution of possible designed capacity of ESS.The design scenarios may be classified as a few different sce-narios, i.e. cheap infrastructure design, moderate infrastruc-ture design, and expensive infrastructure design.4.2 The Effect of Pantograph Voltage Basically,

the variation of pantograph voltage directly affects the per-formance of the traction system. To simplify the problemsolved in the proposed design, the optimization process dis-regarding the voltage effect may provide nonpractical solu-tions. From the comparison between the numerical results



(a) Speed Profile, Pantograph Voltage profile (b) Train’s power, and SOC VS Time

Fig. 11. Speed profile, Pantograph voltage profile, Train’s power, and SOC for designed operating conditionOD4 (Peak hour)

(a) Speed Profile, Pantograph Voltage profile (b) Train’s power, and SOC VS Time

Fig. 12. Speed profile, Pantograph voltage profile, Train’s power, and SOC for designed operating conditionOD7 (Off-peak hour)

(a) %Esave VS Capacity of ESS (Eess) (b) %Ereg VS Capacity of ESS (Eess)

Fig. 13. Comparisons of %Esave, %Ereg, Capacity of ESS (Eess) with and without effect of pantograph voltage

with and without the effect of voltage, the variation of voltagein an acceptable range entails some small errors in the designscheduling that may degrade the energy-saving performance.Therefore, a small deviation in some running sections maybe allowed if the design scheduling and speed profiles areapplied for in the system operation without any modification.4.3 Concerns Regarding Application of the Proposed

Method The integrated design aims to maximize the uti-lization of regenerative energy by combining smart schedul-ing and use of onboard energy storage. Generally, urbanrailways operated in peak hour period or small headway pe-riod have good utilization of regenerative energy. There-fore, the proposed design tends to be ineffective at improv-ing energy-saving performance but reducing the peak power

at a substation may be advantageous. The system operatingin off-peak periods or larger headway periods obviously hasthe possibility to obtain more effective regenerative energymanagement. From the numerical results, the design operat-ing condition provides better improvement on energy-savingperformance when compared with nominal operations.

5. Conclusions

This paper presents the integrated design of train schedul-ing, use of onboard energy storage, and traction power man-agement. The main objective is to combine the design oftrain operation and infrastructure to improve energy-savingoperations and the flexibility of energy management. The de-sign solution is based on minimizing energy supplied from



Fig. 14. Peak power at power substation

substations and the energy capacity of onboard ESS. More-over energy-saving and cost-saving purposes can be compro-mised by varying weighting factor. From the numerical re-sults evaluated in the case studies of the BTS, the designtimetable parameters and appropriate capacity of ESS ob-tained from the proposed design can provide improvement ofenergy-saving performance of up to 9.65% when comparedwith nominal operation and can reduce the peak power at asubstation by approximately 40%. The design scenario canbe simply classified as cheap, moderate, and expensive de-sign depending on variations of the weighting factor.

From the case studies, when peak hour (3-minute head-way) is considered as part of the design condition, utiliz-ing the designed scheduling and increasing ESS’s capacityfrom 1 kWh to 2 kWh provided an improvement of energy-saving performance from 4.12% to 5.87%. In the case ofoff-peak hour (5-minute headway), the energy-saving per-formance can be improved from 5.7% to 9.65% by upgrad-ing ESS capacity from 1 kWh to 3 kWh with the designedscheduling.

Furthermore, the effect of pantograph voltage has also beenevaluated and compared with an ideal case and the results in-dicate that when the effect of voltage is included in the calcu-lation, energy-saving performance is degraded due to somesmall errors in the design scheduling. When the effect ofpantograph voltage is considered, the energy-saving perfor-mance in peak hour and off-peak hour is degraded to 5.30%and 8.88% respectively. The variation of train voltage affectsthe design scheduling and some small deviations in runningtime of some sections may be allowed. The application ofproposed design still provides considerable improvement inenergy-saving operation. From the comparisons, we see ap-proximately 1% of energy-saving is reduced due to the volt-age effect.

References

( 1 ) X. Yang, X. Li, B. Ning, and T. Tang: “A survey on energy-efficient trainoperation for urban rail transit”, IEEE Trans. on Intelligent TransportationSystems, Vol.17, pp.2–13 (2016)

( 2 ) R. Takagi: “Energy saving techniques for the power feeding network ofelectric railways”, IEEJ Trans. Electrical and Electronic Engineering, Vol.5,pp.312–316 (2010)

( 3 ) Z. Tian, S. Hillmansen, C. Roberts, P. Weston, N. Zhao, L. Chen, and M.Chen: “Energy evaluation of the power network of a DC railway system with

regenerating trains”, IET Electrical Systems in Transportation, Vol.6, pp.41–49 (2016)

( 4 ) M. Miyatake and H. Ko: “Optimization of train speed profile for minimumenergy consumption”, IEEJ Trans. Electrical and Electronic Engineering,Vol.5, pp.263–269 (2010)

( 5 ) R. Chevrier, G. Marliere, B. Vulturescu, and J. Rodriguez: “Multi-objectiveevolutionary algorithm for speed tuning optimization with energy saving inrailway: Application and case study”, in RailRome 2011 (2011)

( 6 ) G. Matsuura and M. Miyatake: “Optimal train speed profiles by dynamicprogramming with parallel computing and the fine-tuning of mesh”, WITTransactions on The Built Environment, Vol.135, pp.767–777 (2014)

( 7 ) T. Albrecht: “Reducing power peaks and energy consumption in rail transitsystems by simultaneous train running time control”, WIT Transactions onState-of-the-art in Science and Engineering, Vol.39 (2010)

( 8 ) X. Li and H.K. Lo: “Energy minimization in dynamic train scheduling andcontrol for metro rail operations”, Transportation Research Part B: Method-ological, Vol.70, pp.269–284 (2014)

( 9 ) X. Sun, S. Zhang, H. Dong, Y. Chen, and H. Zhu: “Optimization of metrotrain schedules with a dwell time model using the lagrangian duality theory”,IEEE Transactions on Intelligent Transportation Systems, Vol.16, pp.1285–1293 (2015)

(10) F. Lin, S. Liu, Z. Yang, Y. Zhao, Z. Yang, and H. Sun: “Multi-train energysaving for maximum usage of regenerative energy by dwell time optimizationin urban rail transit using genetic algorithm”, Energies, Vol.9, p.208 (2016)

(11) S.D. Gupta, J.K. Tobin, and L. Pavel: “A Robust Mixed Integer Optimiza-tion Model to Utilize Regenerative Energy of Trains in a Railway Network”,arXiv preprint arXiv:1507.01646 (2015)

(12) L. Battistelli, M. Fantauzzi, D. Iannuzzi, and D. Lauria: “Energy manage-ment of electrified mass transit systems with Energy Storage devices”, inPower Electronics, Electrical Drives, Automation and Motion (SPEEDAM),2012 International Symposium on (2012)

(13) H.-S. Lee, H.-M. Lee, C.-M. Lee, G.-S. Jang, and G.-D. Kim: “Energy stor-age application strategy on DC electric railroad system using a novel railroadanalysis algorithm”, Journal of Electrical Engineering and Technology, Vol.5,pp.228–238 (2010)

(14) B. Jung, H. Kim, H. Kang, and H. Lee: “Development of a Novel Charg-ing Algorithm for On-board ESS in DC Train through Weight Modification”,Journal of Electrical Engineering & Technology, Vol.9, pp.1795–1804 (2014)

(15) T. Ratniyomchai and T. Kulworawanichpong: “A Demonstration Project forInstallation of Battery Energy Storage System in Mass Rapid Transit”, inEnergy Procedia (2017 International Conference on Alternative Energy inDeveloping Countried and Emerging Economies), Bangkok (2017)

(16) T. Ratniyomchai, S. Hillmansen, and P. Tricoli: “Optimal capacity and posi-tioning of stationary supercapacitors for light rail vehicle systems”, in PowerElectronics, Electrical Drives, Automation and Motion (SPEEDAM), 2014International Symposium on (2014)

(17) F. Ciccarelli and D. Iannuzzi: “A Novel energy management control ofwayside Li-Ion capacitors-based energy storage for urban mass transit sys-tems”, in Power Electronics, Electrical Drives, Automation and Motion(SPEEDAM), 2012 International Symposium on (2012)

(18) V. Calderaro, V. Galdi, G. Graber, and A. Piccolo: “Optimal siting and sizingof stationary supercapacitors in a metro network using PSO”, in IndustrialTechnology (ICIT), 2015 IEEE International Conference on (2015)

(19) R. Barrero, X. Tackoen, and J. Van Mierlo: “Stationary or onboard energystorage systems for energy consumption reduction in a metro network”, Pro-ceedings of the Institution of Mechanical Engineers, Part F: Journal of Railand Rapid Transit, Vol.224, pp.207–225 (2010)

(20) V. Gelman: “Energy storage that may be too good to be true: Comparison be-tween wayside storage and reversible thyristor controlled rectifiers for heavyrail”, IEEE Vehicular Technology Magazine, Vol.8, pp.70–80 (2013)

(21) C.H. Bae, M.S. Han, Y.K. Kim, S.Y. Kwon, and H.J. Park: “A study of thepower capacity of regenerative inverters in a DC electric railway system”,Power Supply, Energy Management and Catenary Problems, p.35 (2010)

(22) W. Kampeerawat, T. Koseki, and F. Zhou: “Efficient Urban Railway Designintegrating Train Scheduling, Onboard Energy Storage, and Traction PowerManagement”, 2018 International Power Electronics Conference (IPEC-Niigata 2018 -ECCE Asia), Niigata, Japan, 2018, pp.3257–3264 (2018)

(23) S. Su, T. Tang, X. Li, and Z. Gao: “Optimization of multitrain operations in asubway system”, IEEE Trans. on Intelligent Transportation Systems, Vol.15,No.2, pp.673–684 (2014)

(24) S. Su, T. Tang, and C. Roberts: “A cooperative train control model for energysaving”, IEEE Trans. on Intelligent Transportation Systems, Vol.16, No.2,pp.622–631 (2015)

(25) F. Cao, Z. Liao, S. Liu, J. Xun, and X. Luo: “A two-layer construction ofenergy optimization approach for timetable”, In proceedings of IEEE Inter-national Conference on Intelligent Rail Transportation (ICIRT), pp.476–481



(2016)(26) N. Zhao, C. Roberts, S. Hillmansen, Z. Tian, P. Weston, and L. Chen: “An

integrated metro operation optimization to minimize energy consumption”,Transportation Research Part C: Emerging Technologies, Vol.75, pp.168–182(2017)

(27 ) M. Domınguez, A. Fernandez-Cardador, A.P. Cucala, and R.R. Pecharroman:“Energy savings in metropolitan railway substations through regenerative en-ergy recovery and optimal design of ATO speed profiles”, IEEE Trans. onAutomation Science and Engineering, Vol.9, No.3, pp.496–504 (2012)

(28) W. Kampeerawat and T. Koseki: “A strategy for utilization of regenerativeenergy in urban railway system by application of smart train schedulingand wayside energy storage system”, Energy Procedia, Vol.138, pp.795–800(2017)

(29) W. Kampeerawat and T. Koseki: “Integrated Design of Train Scheduling andWayside Energy Storage for Energy-Saving Urban Railway Operation—ACase Study of Bangkok Rapid Transit System”, IEEJ Technical Meeting,November 30 (2017)

(30) Y. Cai, M.R. Irving, and S.H. Case: “Iterative techniques for the solu-tion of complex DC-rail-traction systems including regenerative braking”,IEE Proceedings-Generation, Transmission and Distribution, Vol.142, No.5,pp.445–452 (1995)

(31) T. Kulworawanichpong: “Multi-train modeling and simulation integratedwith traction power supply solver using simplified Newton–Raphsonmethod”, Journal of Modern Transportation, Vol.23, No.4, pp.241–251(2015)

(32) R. Takagi: “More intelligent DC railway system by integrating electrificationand train group control”,（直流饋電系と列車群制御の統合インテリジェントシステム化）, PhD diss., The University of Tokyo (1995)

(33) MITRAC Energy Saver, Retrieved from https://www.bombardier.com

Warayut Kampeerawat (Student Member) received the B.S. degreeand the M.S. degree in electrical engineering fromKhon Kaen University, Khon Kaen, Thailand, in 2005and 2007, respectively. He is currently pursuing aPh.D. in the Department of Electrical Engineeringand Information Systems, Graduate School of Engi-neering, The University of Tokyo. He is a lecturer inthe Department of Electrical Engineering, Faculty ofEngineering, Khon Kaen University, Thailand. Hisresearch fields include applying optimization tech-

niques to power systems and railway system design.

Takafumi Koseki (Member) received his Ph.D. in electrical engineer-ing from The University of Tokyo, Tokyo, Japan, in1992. He is currently a Professor in the Departmentof Electrical Engineering and Information Systems,School of Engineering, The University of Tokyo. Hiscurrent research interests include applications of elec-trical engineering to public transport systems, lineardrives, and control of traction systems. Dr. Kosekiis a member of the Institute of Electrical Engineers ofJapan, Japan Society of Mechanical Engineering, The

Japan Society of Applied Electronics and Mechanics, Japan Society for Pre-cision Engineering, and Japan Railway Electrical Engineering Association.


Integrated Design of Smart Train Scheduling, Use of ...

Documents

Transcript of Integrated Design of Smart Train Scheduling, Use of ...