An Improved Energy Management Strategy for Hybrid Energy Storage System in Light Rail Vehicles

Transcription

1 eergies Article A Improved Eergy Maagemet Strategy for Hybrid Eergy Storage System i Light Rail Vehicles Log Cheg * ID, Wei Wag, Shaoyua Wei, Hogtao Li ad Zhidog Jia Natioal Active Distributio Network Techology Research Ceter, Beijig Jiaotog Uiversity, Beijig , Chia; wwag2@bjtu.edu.c (W.W.); shaoyuawei@bjtu.edu.c (S.W.); @bjtu.edu.c (H.L.); @bjtu.edu.c (Z.J.) * Correspodece: logcheg1@bjtu.edu.c; Tel.: Received: 17 Jauary 2018; Accepted: 5 February 2018; Published: 12 February 2018 Abstract: A sigle-objective optimizatio eergy maagemet strategy (EMS) for a oboard hybrid eergy storage system (HESS) for light rail (LR) vehicles is proposed. The HESS uses batteries ad supercapacitors (SCs). The mai objective of the proposed optimizatio is to reduce the battery ad SC losses while maitaiig the SC state of charge (SOC) withi specific limits based o the distace betwee cosecutive LR statios. To do this, a series of optimized SOC limits is used to prevet the SC from becomig exhausted prematurely istead of the stadard SC SOC pealty term i the cost fuctio. Meawhile, a rule-based EMS (RB-EMS) is used to give the SCs chargig priority over the batteries whe the vehicle is brakig. Moreover, a simplified method for the optimizatio is proposed to reduce the computatioal burde. Simulatio ad experimetal results for the proposed EMS ad a stadard SC SOC pealty-based cost fuctio optimizatio are provided to evaluate losses. As a result, it is show that the proposed EMS, compared with stadard SC SOC pealty-based cost fuctio optimizatio, decreases losses ad prevets the SOC from reach the dischargig limits. Keywords: eergy maagemet strategy; supercapacitors; batteries; light rail 1. Itroductio A typical light rail (LR) etwork has a high volume, a high desity of vehicle operatio, ad a short distace betwee statios. Most of these LR etworks are powered by a overhead cateary system, which has some egative cosequeces [1]: First, a cateary system visually pollutes a city s ifrastructure. Secod, peak power delivery ad cosumptio patters of the LR vehicle ca provoke power supply issues, such as voltage variatios. Third, the cateary system has a low eergy efficiecy due to limited brakig eergy recovery ad high losses. A effective solutio to the above issues is the implemetatio of a eergy storage system (ESS) o urba trasportatio etworks for brakig eergy recovery purposes [2]. Aother feasible solutio is the direct istallatio of a ESS o LR vehicles that ca meet the requiremets of tractio applicatios ad serve as power supplies [3 6]. A oboard ESS for LR vehicles has the followig advatages: First, urba trasportatio areas ca be exteded to cateary-free zoes. Secod, eergy trasmissio does ot eed to go through the cateary, which sigificatly miimizes trasportatio systems losses [7]. Third, it allows a flexible ad simpler ifrastructure. Fially, the eergy recovery of a oboard ESS, compared to its competitors, has greater flexibility ad lower ivestmet costs [6]. Batteries have bee cosidered the most commo eergy storage compoet of a ESS because of their high eergy desity, compact size, ad reliability [8 10]. However, they also have well-kow disadvatages such as low specific power, poor properties at low temperature, ad a short life-cycle. Supercapacitors (SCs) by compariso have much higher power desities ad much loger service Eergies 2018, 11, 423; doi: /e

2 Eergies 2018, 11, of 15 lives, but lower eergy desities [1]. Additioally, the use of SCs as auxiliary ESSs for hybrid eergy storage systems (HESSs) has bee demostrated to icrease the system s peak power, reduce iteral losses, ad assist batteries durig peak power demads ad regeerative brakig. Therefore, such use reduces battery drai, legthes battery life, ad eables the use of smaller battery packs [9,11]. The HESSs combiig two or more types of storage compoets with complemetary features have bee extesively studied, ad HESSs composed of batteries ad SCs are oe of the most promiet combiatios [9,12 19]. I a HESS, SCs are usually used to absorb the high power of regeerative brakig ad to supply maximum power for acceleratio, while batteries are used for vehicle operatios that ivolves less/rated power [20]. Recetly, the developmet of cotrol strategies for eergy maagemet systems (EMS) has become a topic of iterest for researchers [5,11 13,21 28]. The EMS plays a importat role to esure that the oboard ESS ca meet eergy reductio requiremets [24,26]. Meawhile, it ca also optimize objectives by, for example, costraiig the magitude of the battery charge/discharge curret to exted its life-cycle, miimizig HESS losses, ad maitaiig the state-of-charge (SOC) of SCs i a suitable rage to make sure that it has eough eergy to work. A thorough survey of the curret literature shows that the research works cocerig EMS ca be classified ito two categories. The first category of EMS uses a qualitative rule to split the total required power ito two parts, such as a battery ad a SC, based o several iput parameters, i.e., power demad, maximum battery power, SC SOC ad battery SOC, etc. Amog these EMSs, the most commo method is the rule-based eergy maagemet strategy (RB-EMS) [11,21 23]. This cotrol method is very easy to implemet ad deals with the limitatios of the battery curret referece, the SC power referece, etc. RB-EMS is based o a set of coditios that have bee decided previously. However, it still has two mai drawbacks: First, the distributio rule based o the iput parameters, such as the SC SOC ad the battery SOC, is hard to desig due to the differet power demad coditios that eed to be evaluated. Secod, the HESS loss calculatio is ot cosidered carefully. I this category of EMS, fuzzy logic cotrol [26,27] has bee developed for power sharig i HESS. These fuzzy-logic-based cotrol algorithms ca make this distributio rule more adaptable to complex power demad coditios based o vague or imprecise iformatio [27]. However, the foudatio of the membership fuctio of fuzzy cotrol is still based o experiece istead of careful calculatio. For the RB-EMS, most of the iput parameters of the cotroller eed to be decided without quatitative cosideratio, which meas it is hard to achieve optimal performace. The secod category is to formulate the HESS power sharig problem as a covex optimizatio problem that miimizes the system losses (as a major objective) with a cost fuctio. I [13,24,25,28], optimizatio algorithms ad model predictive cotrol (MPC) are proposed to achieve the goal of reducig system losses, employig loss calculatios of curret istats or a predictio horizo with discrete models. However, these EMSs also have two mai drawbacks: First, predictio horizos are relatively short (less tha 3 i these papers). As a result, the optimizatio horizo eeds to be tued i the simulatio for a better performace [28]. Secod, a sigle-objective cost fuctio that miimizes power losses caot be directly implemeted. The mai reaso is that the eergy provided by the SC is cheaper tha the battery i terms of losses, so the SC will be discharged to its miimum SOC limit prematurely. To solve this problem, all of the above-metioed literature employs a multi-objective cost fuctio that ot oly takes power loss ito accout but also pealizes deviatio from the SC SOC operatig poit. However, this solutio also eeds to tue the pealty weights to improve performace, especially based o differet power demad profiles. For istace, as described i [28], to further utilize the SC SOC rage, parameters i the pealty fuctio eed to be tued. Meawhile, the multi-objective cost fuctio causes the eergy distributio result to deviate from the origial sigle objective, miimizig power loss, which is a drawback. The aim of this work is to propose a EMS based o a sigle-objective optimizatio strategy that miimizes the total losses of the battery ad the SC. Meawhile, a variable-legth predictio horizo based o a slidig forward widow strategy is proposed to exted the optimizatio horizo

3 Eergies 2018, 11, of 15 to cover the etire LR tractio process. Moreover, the pealty term based o the deviatio from the SC SOC operatig poit is replaced with a SC SOC limit based o the distace betwee cosecutive LR statios. A RB-EMS is used to preferetially charge the SC. As a result, a sufficietly log predictio horizo optimizatio based o the HESS losses is achieved i each etire LR tractio power profile. Lastly, simulatio ad experimetal results are provided to evaluate the proposed EMS i terms of loss ad SC SOC limit objectives. The performace is also compared with a stadard SC SOC pealty-based cost fuctio optimizatio. 2. Oboard HESS Cofiguratios Aalysis Various topologies for HESSs combiig batteries ad SCs have bee proposed [29 31]. I geeral, oboard ESSs cosist of a DC mai ESS ad oe or more DC/DC coverters. With such cofiguratio, the use of DC/DC coverters is essetial for cotrollig power flow at differet voltage levels. I Figure 1a,b, the SC bak ad battery bak are coected i parallel with, ad through, a DC/DC coverter or directly to the DC bus [29]. This cofiguratio is the simplest battery/sc HESS combiatio, but the power distributio betwee the batteries ad the SCs is ucotrolled. I additio, the coclusios of [30] idicate that a battery SC hybrid power source ca supply a pulsed load with higher peak power, but due to the limitatio of the battery bak voltage to the SC bak, the curret supplied by the SC bak drops very quickly. Ultimately, the battery SC hybrid power oly shows the battery characteristics, which caot reflect the power characteristics of the SC bak. Battery DC bus Iverter Motor Battery DC bus DC-DC Iverter Motor Super Capacitor Battery (a) DC bus DC-DC Iverter Motor Super Capacitor Battery (d) DC bus DC-DC Iverter Motor Super Capacitor Super Capacitor (b) DC bus DC-DC Iverter Motor Super Capacitor DC-DC (e) Battery (c) Figure 1. Oboard eergy storage system (ESS) cofiguratios: (a) Passive battery/sc (supercapacitor). (b) Passive cascaded battery/sc. (c) Active cascaded SC/battery. (d) Active cascaded battery/sc. (e) Multiple DC/DC coverters. A widely used cofiguratio of hybrid HESS i HEVs is show i Figure 1c [31]. The DC/DC coverter allows the SC to be used o its wide voltage rage ad the battery bak is directly coected to the DC bus, while maitaiig a relatively costat dc-lik voltage. But the power of the DC/DC coverter must match the large ad itermittet SC power, while the battery is forced to suffer from a high frequecy curret. I Figure 1d, the positios of the battery bak ad SC bak are reversed with respect to Figure 1c. I this way, the DC/DC coverter ca be smaller ad the battery curret ca be cotrolled, but the DC bus voltage will fluctuate i a wide rage because of the ucotrolled workig voltage rage of the SC. Fially, i Figure 1e, there are two DC/DC coverters for the battery bak ad the SC bak, respectively. I this cofiguratio, all mai electrical variables, i.e., the battery curret, the SC curret, ad the DC bus voltage ca be cotrolled. The mai disadvatage is that two

4 Eergies 2018, 11, of 15 coverters are required, which icreases size ad cost. To cotrol the flexibility of both eergy storage compoets ad the DC bus voltage stability, this topology is used i this work. 3. Eergy Maagemet Strategy Desig 3.1. Descriptio This work proposes a sigle-objective optimizatio EMS to miimize the battery ad SC losses. The mai ideal is to formulate a cost fuctio based o a variable-legth predictio horizo to optimize the power sharig problem betwee cosecutive LR statios. Meawhile, the SOC limitatios of SC are obtaied due to the predictio horizo. I this way, a series of SOC limitatios based o optimizatios i each tractio/brake stage, istead of the pealty term i the cost fuctio, is used to prevet the SC from becomig exhausted prematurely Proposed Slidig Forward Widow Strategy i Stage Basically, the etire power demad profile of a LR route from the first to the last statio cosists of a series of iter-statio power demad stages. Each stage has two parts: a tractio stage ad a brakig stage. I the tractio stage, the eergy is provided by a HESS to meet the eeds of the LR acceleratio operatio ad is the trasferred back to the HESS i the brakig stage to improve the efficiecy. As described i [26], the power demad profile of a LR route, ulike that of electric vehicles, is kow at the begiig of the jourey, which makes the future power demad kow at each istat k i the predictio horizo. Based o this, a slidig forward widow strategy i a LR tractio stage is proposed. As show i Figure 2, a LR tractio stage is divided ito a series of steps. The variable-legth predictio horizos will exted to the ed of this tractio stage i each step. For istace, the predictio horizo is i at step i, ad it is chaged to j at step j. I this way, the slidig forward widow will move as the predictio step icreases to always cover the etire tractio stage. The proposed EMS cosiders ot oly a few steps i the future but also the etire LR tractio stage horizo to miimize battery ad SC losses, so that it ca avoid fallig ito the zoe that deviates from the optimal poit, such as whe the SC is exhausted prematurely ad the batteries the have to supply the power demad aloe, which icreases the total loss. i i i 1 Power demad j k i i+1i+2... j j+1 j+2... k k+1... Predictio steps Figure 2. Block diagram of the proposed slidig forward widow strategy i the tractio stage.

5 Eergies 2018, 11, of Proposed EMS i the Brakig Stage I order to miimize the battery ad SC losses, the SC should be used as much as possible durig the tractio stages. To achieve this, it is reasoable to give SC chargig priority over the battery i the brakig stages. As show i Equatio (1), a RB-EMS is used i brakig stages to split the power to charge both the battery ad the SC. P d (k) P d (k) Pdcdc max ad SOC SC(k) < SOC SC,max P SC (k) = Pdcdc max P d (k) > Pdcdc max ad SOC SC(k) < SOC SC,max 0 SOC SC (k) SOC SC,max P d (k) P SC (k) SOC bat (k) < SOC bat,max P bat (k) = 0 SOC bat (k) SOC bat,max (1) where P SC ad P bat represet the chargig power of the SC ad the battery, respectively. The P d refers to the power demad of the LR i the brakig stage, Pdcdc max is the maximum power of the DC-DC coverter, ad SOC SC,max ad SOC bat,max are the maximum SOC of the battery ad the SC, respectively SC Costraits i Each Stage A practical LR system i Zhuhai, Chia, is used to show how the proposed EMS works. The LR power demad ad speed profile i the first three stops are show i Figure 3. The speed curve is cosidered as a startig poit for the desig of the LR system by the maufacturer, ad the power curve is calculated based o the speed curve ad the parameters of the LR. For the auxiliary power load, a average value of 55 kw is cosidered durig the etire lie. [km/h] Brakig Brakig Brakig (a) (b) [kw] Figure 3. 3 Typical stops i the drivig cycle of the Light Rail (LR): (a) speed ad (b) power demad profile. The power demad profile i Figure 3 is divided ito ie stages alterately: three tractio stages, three brakig stages, ad three stop stages. I each tractio stage, the proposed EMS splits the power demad of the battery ad the SC by formulatig a cost fuctio that miimizes their losses. Sice each LR tractio stage predictio horizo formulatio is used ad the SC costs less tha batteries i term of losses, the miimum SOC limitatio of the SC is probably the ed SOC of the SC i each tractio stage. Therefore, each tractio stage of the SC has a miimum SOC costrait at the ed so that the SC is ot over-discharged too early, which has the same effect as pealizig the deviatio from the SC SOC operatig poit referece ito the cost fuctio. Whe the LR is brakig, the proposed RB-EMS, which gives chargig priority to the SC, is used, ad the SC will be charged if there is a charger i the substatios.

6 Eergies 2018, 11, of Battery Model ad Super Capacitor Model For this case, a Toshiba 8.5 Ah Lithium titaate battery was chose. This battery has better power specificatios [32] ad a wider operatig temperature rage compared with other batteries of the same type ad is more suitable for applyig to LR vehicles [33]. A Thevei battery model was used for the battery, ad the model parameters were obtaied by idetificatio techiques, as described i [34]. Figure 4a shows the battery circuit model, where V ocv is the battery ope-circuit voltage (OCV) ad R bat is the battery iteral resistace. The parameters C p ad R p model the trasiet voltage respose of the battery, ad V p represets their voltage. These parameters V ocv, R o, C p, ad R p are all variables based o differet battery SOCs as show i Figure 5. The term i o represets the battery output curret. The power loss of the battery at ay time k ca be expressed as Loss bat (k) = i 2 o(k)r bat (k). (2) + - V ocv R bat + V p - C p R p i o + V bat V ocv R SC i o + V SC - (a) (b) Figure 4. Circuit diagram of (a) the battery model ad (b) the SC model. Vocv [V] R o chargig R o dischargig V ocv chargig V ocv dischargig Ro [mω] SOC[%] Figure 5. Parameters V ocv ad R o i the 8.5 Ah Thevei battery circuit model. Figure 4b shows the SC model, where V ocv is the SC OCV, ad R SC is the iteral resistace. The value of R SC ca be obtaied from the maufacturer s data sheet as a costat value, ad the SC s power losses expressio is same as Equatio (2), but cosiderig its iteral resistace as Loss SC (k) = i 2 o(k)r SC (k). (3)

7 Eergies 2018, 11, of The Proposed Cost Fuctio ad Costraits I each tractio stage, the cost fuctio that miimizes the HESS losses based o Equatio (2) ad Equatio (3) is writte as follows: E Loss = i=1 Loss bat (k) = Loss SC (k) = α(k) [0, 1] P Loss (k)t s = ( Loss bat (k)+ ( P d(k)[1 α(k)] ) 2 R V bat (k) bat (k)t s ( P d(k)α(k) V SC (k) )2 R SC (k)t s Loss SC (k))t s (4) where E Loss ad P Loss (k) represet the total loss eergy of the HESS i a tractio stage ad the power of loss at istat T s, respectively, P d (k) ad α(k) refer to the power demad of the LR vehicle ad the power mix to split the power demad betwee the battery ad the SC at istat k, respectively, ad Loss bat (k) ad Loss SC (k) are the power of loss from the battery ad the SC at istat k, respectively. Thus, the predictio of the SC ad the battery voltage i k+ 1 based o the models ca be obtaied as V SC (k+1) = VSC 2 (k) 2P d(k+1)α(k+1)t s C V bat (k+1) = V ocv (k) P d(k+1)r bat (k)(1 α(k+1)) V p (k) V bat (k) V p (k) = V p (k 1)+(i bat (k 1) V p(k 1) R p (k 1) ) T s C p (k 1). Based o Equatios (4) ad (5), the total loss of HESS E Loss i a tractio stage ca be expressed via (5) E Loss = ( P d(k)[1 α(k)] ) 2 R V bat (k) bat T s + SOC bat,mi SOC bat (k) SOC bat,max SOC SC,mi SOC SC (k) SOC SC,max [P d (k)α(k)] 2 R SC T s (VSC 2 (1) k 1 P d (i)α(i)) i=1 (6) C Bat (k) C Bat,max 0 α(k) 1 where SOC bat,mi, SOC bat,max, SOC SC,mi, ad SOC SC,max are the SOC limits of the battery ad the SC, respectively. C Bat,max is the maximum curret rate of the battery. The cost fuctio, Equatio (6), is miimized by usig MATLAB sfmico fuctio [28], maily with the SC SOC costrait SOC SC,mi as a fial poit, 25%, where a SC releases 75% of its stored eergy [25]. Meawhile, iitializatio strategies ca be used to speed up the optimizatio computatio as described i [35]. Accordig to the total time of oe tractio stage, as show i Figure 3, here the T s is set to be 2 s. Cosiderig that the DC-DC coverter loss calculatio is always a kid of approximate calculatio [28] ad that this study is iteded to simulate the power flow betwee the battery ad the SC, the DC-DC coverters losses are cosidered costat. Oly the losses of the eergy storage compoets (the battery ad the SC) are cosidered i this work Simplified Predictio Horizo ad Miimizatio It is well kow that log horizo predictios icrease computatioal burde. However, a movig block techique, where the predictio horizo,, is divided ito two parts = 1 + 2, ad

8 Eergies 2018, 11, of 15 a larger samplig period for the secod horizo achievig 2 to obtai less computatioal complexity has bee adopted i [36]. Followig this method, here a variable-horizo ca be used to decrease the computatioal burde with a relatively iaccurate miimum loss based o a larger samplig period i 2. The above simplified solutio ca reduce the predictio horizo but relatively icrease the loss of the HESS. The performace of this solutio will be discussed i the followig sectio. 4. Simulatio Results To demostrate the performace of the proposed EMS, simulatios were performed i MATLAB/Simulik. The first three stops of the power demad profile are show i Figure 3b. Cosiderig that the total route has 20 stops ad the total eergy demad is 60 kwh, here the sizig of the battery is set at 78 kwh to provide additioal eergy to overcome the capacity declie. The SC sizig is set at 5.3 kwh to help the battery to supply the maximum power demad poit. Based o the total sizig of the battery ad SC ad the voltage level of the HESS (approximate 500 V), the umbers of series ad parallel are obtaied. The battery ad SC sizig parameters are show i Tables 1 ad 2 respectively. The power mix α is optimized at every 2 s. Table 1. Battery parameters. Parameter Descriptio Value m Number of cells i series 200 Number of cells i parallel 20 C [Ah] Rated Capacitace 8.5 V out [V] Rated Voltage 2.3 I max [A] Charge/discharge curret 85 (10 C) Table 2. Supercapacitor parameters. Parameter Descriptio Value m Number of modules i series 10 Number of modules i parallel 20 C [F] Rated Capacitace 83 V out [V] Rated Voltage 48 I max [A] Absolute Maximum Curret 1150 R o [mω] Maximum ESR 10 The results with two differet cost fuctios are show i Figures 6 ad 7. As show i Figure 6b, the SC reaches its miimum SOC limit, 240 V, whe the trai does ot fiish the last two tractio stages with oe-step predictive horizo sigle-objective optimizatio EMS. The battery the has to supply the power demad aloe from t = 103 to t = 130 s ad from t = 205 to t = 223 s, which is ot ideal. [A] i bat V 400 bat VSC i SC (a) (b) [V] Figure 6. Simulatio results for oe-step predictive horizo sigle-objective optimizatio eergy maagemet systems (EMS with three typical stops i the drivig cycle of the LR: (a) battery curret ad SC curret ad (b) battery voltage ad SC voltage.

9 Eergies 2018, 11, of 15 [A] i bat V 400 bat VSC i SC (a) (b) [V] Figure 7. Simulatio results for MPC based two-objectives cost fuctio optimizatio eergy maagemet systems (EMS) with three typical stops i the drivig cycle of the LR: (a) battery curret ad SC curret ad (b) battery voltage ad SC voltage. I Figure 7, the MPC-based cost fuctio has two terms: miimizig the loss ad pealizatio for SC SOC. The pealizatio term is defied as follows [28]: f sc (k) = 1+ sg(i sc (k)(1 e i sc(k) )( SOC sc,mid S0C sc (k) SOC sc,mid S0C sc,mi ) (7) where SOC sc,mid is the selected SC SOC midpoit, SOC sc,mi is the SOC limit of the SC, ad i sc (k) is the SC output curret. It ca be see that the SC works durig all tractio stages due to the pealizatio term, ulike i the oe-step predictive horizo sigle-objective optimizatio EMS. The total loss of the battery ad the SC with the oe-step predictive horizo optimizatio EMS ad the SC SOC pealty MPC-based optimizatio EMS are kwh ad kwh, respectively. I Figure 8 the proposed slidig forward widow optimizatio EMS allows the SC SOC to reach the miimum SOC limit, 240 V, at the ed of the last two tractio stages whe t = 130 s ad t = 223 s. The total loss of the battery ad the SC with the proposed EMS is kwh. The simulatio results show that the proposed eergy maagemet strategy has a lower loss compared with the MPC-based two objectives cost fuctio optimizatio. [A] i bat V 400 bat VSC i SC (a) (b) [V] Figure 8. Simulatio results for the proposed slidig forward widow optimizatio EMS with three typical stops i the drivig cycle of the LR: (a) battery curret ad SC curret ad (b) battery voltage ad SC voltage. Figure 9 shows the simplified method as described i Sectio 3.7. I Figure 9c, the predictio iterval is chaged i each tractio stage. For istace, i Areas, A, B, ad C, the power mix α is optimized every 2 s, every 4 s, ad oly oce, respectively, i each area. The total loss of the battery ad the SC with the variable predictive horizo optimizatio EMS is kwh, but it is oly 1.5% higher tha the loss with the proposed slidig forward widow optimizatio EMS.

10 Eergies 2018, 11, of 15 [A] (a) [%] i bat [V] i SC A B C (c) α V bat (b) 250 VSC Figure 9. Simulatio results for variable predictive horizo optimizatio EMS with three typical stops i the drivig cycle of the LR: (a) battery curret ad SC curret ad (b) battery voltage ad SC voltage ad (c) power mix α. 5. Experimetal Results This sectio presets a experimetal verificatio of the proposed EMS. Three buck/boost bidirectioal coverter circuits were built usig IRF650B MOSFETs, a IFR18650P-1100mAh LiFePO4 battery pack, ad a ZNP-2R7-V-107SS2245 SC pack as iput sources ad a DC bus i parallel as outputs. The first coverter was cotrolled as a reduced power source to simulate the trai power requiremets. The iput curret referece i L1 (k) was set as i L1 (k) = βp d(k)/v i1 (k) (8) where P d (k) is the power demad of the LR, β is the reduced ratio of the power demad profile i Figure 3b, ad v i1 (k) stads for the iput battery pack voltage of the first coverter. The other two coverters are cotrolled uder DC bus voltage cotrol ad iput curret cotrol, respectively, to simulate the HESS. The secod coverter, with aother LiFePO4 battery pack as a iput source, cotrols the DC BUS voltage with a outer voltage PI cotroller ad a ier curret PI cotroller. The third coverter employs the SC pack to cotrol the SC curret. The iput SC curret referece i L3 (k) is set as i L3 (k) = βα(k)p d(k) (9) where α(k) is the power mix. As described i [26], the power mix α(k) ca be obtaied by iterpolatig the LR traveled distace (usig look-up-tables), but here it is obtaied with a time couter i the cotroller usig a look-up-table, which was obtaied i simulatio. The cotroller was implemeted o a dspace DS1005 ad a Slave DSP PWM Geeratio to geerate three 20 khz symmetrical PWMs. The experimetal platform parameters are show i Table 3 ad the setup is show i Figure 10. As show i Figure 11, the SC voltage reaches its limit of 6 V whe t = 272 s; the battery the has to supply the power demad aloe from t = 272 to t = 290 s. Thus, the battery curret reaches 1.6 A. I Figure 12, with the two-objectives MPC-based cost fuctio (miimizig the loss ad pealizatio for SC SOC), the SC works durig the power demad, which is a positive outcome. However, without a SC SOC limit, the SC voltage reaches 6.5 V ad 4.5 V whe the LR fiishes its last two tractio stages. Cosiderig that the battery ad SC losses caot be measured i real systems, here the losses are calculated i the simulatio with these experimetal parameters. The total

11 Eergies 2018, 11, of 15 loss of the battery ad SC with the oe-step predictive horizo optimizatio EMS ad the SC SOC pealty-based optimizatio EMS are W ad W. Table 3. Experimetal platform parameters. Parameter Descriptio Value m bat Number of battery cells i series 4 bat Number of battery cells i parallel 1 C bat [Ah] Battery cell rated Capacitace 1.1 V bat [V] Battery cell rated Voltage 3.2 m SC Number of supercapacitor (SC) cells i series 9 SC Number of SC cells i parallel 1 C SC [F] SC cell rated Capacitace 100 V SC [V] SC cell rated Voltage 2.7 L [mh] Iductor of the coverters 450 C f [mf] Capacitor of the coverters 220 f s [KHz] Switchig frequecy 20 v bus [V] Bus voltage referece 20 T s [s] Optimizatio step of the power mix 2 β Power demad reduced ratio F E C D C C A B G Figure 10. Experimetal setup. A is the LiFePO4 battery pack, B is the SC pack, C is the buck/boost coverter, D is the sesor board, E is the dspace board, F is the simulatio computer, ad G is the power of the sesor board ad drive board.

12 Eergies 2018, 11, of 15 12V V SC 10V i SC 1.6A 1A i bat 40s Figure 11. Experimetal results for oe-step predictive horizo sigle-objective optimizatio EMS with three typical stops i the drivig cycle of the LR: SC curret, SC voltage, ad battery curret. 12V V SC 6.5V 4.5V 10V i SC -3.4A 1A i bat 40s Figure 12. Experimetal results for MPC-based two-objectives cost fuctio optimizatio EMS with three typical stops i the drivig cycle of the LR: SC curret, SC voltage, ad battery curret. I Figure 13, the proposed slidig forward widow optimizatio EMS allows the SC SOC to reach 6.0 V ad 5.8 V i the ed of the last two tractio stages whe t = 192 s ad t = 288 s. The last SC voltage is lower tha 6.0 V because the iductor resistace is ot cosidered i the optimizatio. The total loss of the battery ad the SC with the proposed EMS is W. The experimetal results show that the proposed eergy maagemet strategy has a lower loss compared with the pealty-based cost fuctio optimizatio, as also show i the simulatio results.

13 Eergies 2018, 11, of 15 12V V SC 6.0V 5.8V 10V i SC -2.4A 1A i bat 40s Figure 13. Experimetal results for the proposed slidig forward widow optimizatio EMS with three typical stops i the drivig cycle of the LR: SC curret, SC voltage, ad battery curret. 6. Coclusios A oboard ESS as a power supply for LR vehicles ca reder urba trasportatio areas cateary-free zoes ad miimize the effect of cateary systems losses. To combie high eergy desity with high power desity, a HESS with a battery ad a SC is proposed i this paper. Differet eergy maagemet strategies have bee compared i other studies, ad it has bee show that a multi-objective cost fuctio that cosiders both the power loss ad the deviatio from a SC SOC operatig poit eeds to tue the pealty weights. The eergy distributio result deviates from the origial sigle objective, miimizig the power loss, which is a drawback. To solve this problem, this paper proposes a sigle-objective optimizatio EMS for a oboard HESS for light rail vehicles. The mai objective is to reduce battery ad SC losses, while the SC SOC withi specifics limits based o a slidig forward widow is maitaied i each tractio stage, ad to charge the SC as much as possible based o a RB-EMS at each brakig stage. A series of SOC limits based o the optimizatio at each tractio stage, istead of the pealty term i the cost fuctio, is used here to prevet the SC from becomig exhausted prematurely. Moreover, a simplified method for this optimizatio is proposed to reduce computatioal complexity. Simulatio ad experimetal results are used to verify that the proposed EMS has a 7.5% lower loss compared with the stadard SC SOC pealty-based cost fuctio optimizatio ad that the SC SOC always reaches its limitatio by the ed of each tractio stage. Ackowledgmets: This work was supported by the Natioal Key R&D Program of Chia (Grat Number 2017YFB ). Author Cotributios: Log Cheg ad Wei Wag proposed the research topic ad basic idea; Shaoyua Wei coceived ad desiged the experimets; Log Cheg drafted the mauscript; Hogtao Li ad Zhidog Jia performed the experimets. All authors cotributed to the writig of the mauscript, ad have read ad approved the fial mauscript. Coflicts of Iterest: The authors declare o coflict of iterest.

14 Eergies 2018, 11, of 15 Refereces 1. Lhomme, W.; Delarue, P.; Barrade, P.; Bouscayrol, A. Desig ad Cotrol of a supercapacitor storage system for tractio applicatios. I Proceedigs of the Idustry Applicatios Coferece, Kowloo, Hog Kog, Chia, 2 6 October 2005; pp Barrero, R.; Va Mierlo, J.; Tackoe, X. Eergy savigs i public trasport. IEEE Veh. Techol. Mag. 2008, 3, Mir, L.; Etxeberria-Otadui, I.; Areaza, I.P.D.; Sarasola, I.; Nieva, T. A supercapacitor based light rail vehicle: System desig ad operatios modes. I Proceedigs of the 2009 IEEE Eergy Coversio Cogress ad Expositio, Sa Jose, CA, USA, September 2009; pp Garcia, P.; Feradez, L.M.; Garcia, C.A.; Jurado, F. Eergy maagemet system of fuel-cell-battery hybrid tramway. IEEE Tras. Id. Electro. 2010, 57, Torreglosa, J.P.; Garcia, P.; Feradez, L.M.; Jurado, F. Predictive cotrol for the eergy maagemet of a fuel-cell battery supercapacitor tramway. IEEE Tras. Id. Iform. 2013, 10, Arboleya, P.; Bidagure, P.; Armedariz, U. Eergy is o board: Eergy storage ad other alteratives i moder light railways. IEEE Electrif. Mag. 2016, 4, Cheg, L.; Acua, P.; Wei, S.; Fletcher, J.; Wag, W.; Jiag, J. Fast-Swap Chargig: A Improved Operatio Mode For Cateary-Free Light Rail Networks. IEEE Tras. Veh. Techol. 2017, doi: /tvt Tie, S.F.; Ta, C.W. A review of eergy sources ad eergy maagemet system i electric vehicles. Reew. Sustai. Eergy Rev. 2013, 20, Khaligh, A.; Li, Z. Battery, ultracapacitor, fuel cell, ad hybrid eergy storage systems for electric, hybrid electric, fuel cell, ad plug-i hybrid electric vehicles: State of the art. IEEE Tras. Veh. Techol. 2010, 59, Hossai, E.; Perez, R.; Bayidir, R. Implemetatio of hybrid eergy storage systems to compesate microgrid istability i the presece of costat power loads. I Proceedigs of the IEEE Iteratioal Coferece o Reewable Eergy Research ad Applicatios, Sa Diego, CA, USA, 5 8 November 2017; pp Herrera, V.I.; Gaztañaga, H.; Milo, A.; Saez-De-Ibarra, A.; Etxeberria-Otadui, I.; Nieva, T. Optimal eergy maagemet ad sizig of a battery supercapacitor-based light rail vehicle with a multiobjective approach. IEEE Tras. Id. Appl. 2016, 52, Ami; Bambag, R.T.; Rohma, A.S.; Drokers, C.J. Eergy maagemet of fuel cell/battery/ supercapacitor hybrid power sources usig model predictive cotrol. IEEE Tras. Id. Iform. 2014, 10, Zhao, C.; Yi, H.; Yag, Z.; Ma, C. Equivalet series resistace-based eergy loss aalysis of a battery semiactive hybrid eergy storage system. IEEE Tras. Eergy Covers. 2015, 30, Tai, A.; Camara, M.B.; Dakyo, B. Eergy maagemet based o frequecy approach for hybrid electric vehicle applicatios: Fuel-cell/lithium-battery ad ultracapacitors. IEEE Tras. Veh. Techol. 2012, 61, Lu, S.; Corzie, K.A.; Ferdowsi, M. A ew battery/ultracapacitor eergy storage system desig ad its motor drive itegratio for hybrid electric vehicles. IEEE Tras. Veh. Techol. 2007, 56, Amjadi, Z.; Williamso, S.S. Modelig, simulatio, ad cotrol of a advaced luo coverter for plug-i hybrid electric vehicle eergy-storage system. IEEE Tras. Veh. Techol. 2011, 60, Wag, L.; Collis, E.G.; Li, H. Optimal desig ad real-time cotrol for eergy maagemet i electric vehicles. IEEE Tras. Veh. Techol. 2011, 60, Qu, X.; Wag, Q.; Yu, Y.B. Power demad aalysis ad performace estimatio for active-combiatio eergy storage system used i hybrid electric vehicles. IEEE Tras. Veh. Techol. 2014, 63, Saeger, P.; Devillers, N.; Deschikel, K.; Pera, M.C.; Couturier, R.; Gusti, F. Optimizatio of electrical eergy storage system sizig for a accurate eergy maagemet i a aircraft. IEEE Tras. Veh. Techol. 2017, 66, Xiag, C.; Wag, Y.; Hu, S.; Wag, W. A ew topology ad cotrol strategy for a hybrid battery-ultracapacitor eergy storage system. Eergies 2014, 7,

15 Eergies 2018, 11, of Carter, R.; Crude, A.; Hall, P. J. Optimizig for efficiecy or battery life i a battery/supercapacitor electric vehicle. IEEE Tras. Veh. Techol. 2012, 61, Sog, Z.; Hofma, H.; Li, J.; Hou, J.; Ha, X.; Ouyag, M. Eergy maagemet strategies compariso for electric vehicles with hybrid eergy storage system. Appl. Eergy 2014, 134, Trovao, J.P.; Pereiriha, P.G.; Jorge, H.M.; Atues, C.H. A multi-level eergy maagemet system for multi-source electric vehicles A itegrated rule-based meta-heuristic approach. Appl. Eergy 2013, 105, Choi, M.E.; Lee, J.S.; Seo, S.W. Real-time optimizatio for power maagemet systems of a battery/ supercapacitor hybrid eergy storage system i electric vehicles. IEEE Tras. Veh. Techol. 2014, 63, Greewell, W.; Vahidi, A. Predictive cotrol of voltage ad curret i a fuel cell ultracapacitor hybrid. IEEE Tras. Id. Electro. 2010, 57, Herrera, V.; Milo, A.; Gaztañaga, H.; Etxeberria-Otadui, I.; Villarreal, I.; Camblog, H. Adaptive eergy maagemet strategy ad optimal sizig applied o a battery-supercapacitor based tramway. Appl. Eergy 2016, 169, Li, Q.; Che, W.; Liu, Z.; Li, M.; Ma, L. Developmet of eergy maagemet system based o a power sharig strategy for a fuel cell-battery-supercapacitor hybrid tramway. J. Power Sources 2015, 279, Laldi, O.; Moshirvaziri, M.; Trescases, O. Predictive algorithm for optimizig power flow i hybrid ultracapacitor/battery storage systems for light electric vehicles. IEEE Tras. Power Electro. 2013, 28, Pay, S.; Baghzouz, Y. Effectiveess of battery-supercapacitor combiatio i electric vehicles. I Proceedigs of the Power Tech Coferece, Bologa, Italy, July Dougal, R.A.; Liu, S.; White, R.E. Power ad life extesio of battery-ultracapacitor hybrids. IEEE Tras. Compo. Packag. Techol. 2002, 25, Camara, M.B.; Gualous, H.; Gusti, F.; Bertho, A. Desig ad ew cotrol of dc/dc coverters to share eergy betwee supercapacitors ad batteries i hybrid vehicles. IEEE Tras. Veh. Techol. 2008, 57, We, Y.L.; Aziz, M.J.A.; Idris, N.R.N. Modellig of lithium-titaate battery with ambiet temperature effect for charger desig. IET Power Electro. 2016, 9, Liu, S.; Jiag, J.; Shi, W.; Ma, Z.; Guo, H. State of charge ad peak power estimatio of NCM/Li4Ti5O12 battery usig ic curve for rail tractor applicatio. I Proceedigs of the 2014 IEEE Coferece ad Expo Trasportatio Electrificatio Asia-Pacific (ITEC Asia-Pacific), Beijig, Chia, 31 August 3 September Liu, S.; Jiag, J.; Shi, W.; Ma, Z.; Wag, L.Y.; Guo, H. Butler volmer-equatio-based electrical model for high-power lithium titaate batteries used i electric vehicles. IEEE Tras. Id. Electro. 2015, 62, Safdarejad, S.M.; Hedegre, J.D.; Lewis, N.R.; Haseltie, E.L. Iitializatio strategies for optimizatio of dyamic systems. Comput. Chem. Eg. 2015, 78, Karamaakos, P.; Geyer, T.; Maias, S. Direct voltage cotrol of dc dc boost coverters usig eumeratio-based model predictive cotrol. IEEE Tras. Id. Electro. 2013, 29, c 2018 by the authors. Licesee MDPI, Basel, Switzerlad. This article is a ope access article distributed uder the terms ad coditios of the Creative Commos Attributio (CC BY) licese (