Optimal Control Strategy for PHEVs Using Prediction of Future Driving Schedule

Transcription

1 World Elecric Vehicle Journal Vol. 5 - ISSN WEVA Page 149 EVS26 Los Angeles, California, May 6 9, 212 Opimal Conrol Sraegy for PHEVs Using Predicion of Fuure Driving Schedule Daeheung Lee 1, Suk Won Cha 1, Aymeric Rousseau 2, Namwook Kim 2 1 Seoul Naional Universiy, Seoul, Souh Korea, ideal2@snu.ac.kr 2 Argonne Naional Laboraory, Argonne, Illinois, USA Absrac Opimizaion-based conrol mehods for plug-in hybrid elecric vehicles require knowledge abou an enire driving cycle and an elevaion profile o obain opimal performance over a fixed driving roue. This paper deails our invesigaion ino he mehod of using raffic informaion o predic he fuure driving cycle, as well as an examinaion of he opimal conrol sraegy based on Ponryagin s Minimum Principle, in order o minimize fuel consumpion on a given rip disance and o develop a real-ime implemenable conrol sraegy. To predic fuure driving paerns, he Dynamic Programming heory is proposed for he calculaion of vehicle speed wih respec o driving disance, under he assumpion ha daa abou raffic condiions are obained from exernal raffic informaion, such as Inelligen Transporaion Sysems. Predicion of fuure driving speed is achieved by minimizing he proposed cos funcion on each segmen. The resuls of he generaed speed profile can properly esimae he driving paern of he driver. Also, a cosae generaion algorihm is applied o deermine he parameers wih respec o he required power deduced from he prediced driving cycle. The proposed co-sae generaion model can find he esimaed iniial co-sae ha is similar o he opimal co-sae. Simulaion resuls indicae ha his approach guaranees he bes efficiency under reasonable condiions and he minimizaion of fuel consumpion on he rip disance beween he origin and desinaion. Keywords: PHEV (plug-in hybrid elecric vehicle), EREV (exended range elecric vehicle), power managemen, conrol sysem, ciy raffic 1 Inroducion Plug-in hybrid elecric vehicles (PHEVs) have become an effecive soluion o mee he need for more fuel-efficien vehicles and o address ighening emission regulaions. The PHEVs are funcionally similar o convenional hybrid elecric vehicles (HEVs), since hey can ake advanage of regeneraive braking and a reducion in engine size o operae more efficienly. However, PHEVs differ from convenional HEVs in ha hey allow recharging from he elecrical grid. In hese vehicles, higher energy eries can be recharged from he elecrical oule in addiion o he engine and regeneraive braking. By using a high-energy ery, PHEVs can ravel in pure elecric mode for a specific disance according o he operaing mode of he powerrain [1, 2]. The performance or efficiency of PHEVs also relies on is vehicle energy managemen sraegy, EVS26 Inernaional Baery, Hybrid and Fuel Cell Elecric Vehicle Symposium 1

2 World Elecric Vehicle Journal Vol. 5 - ISSN WEVA Page 15 which deermines he power-spli raio beween he combusion engine and he elecric diving moor o mee he required power a he driving wheels [3]. Opimizaion-based conrol mehods, such as he Equivalen Consumpion Minimizaion Sraegy (ECMS) or Ponryagin s Minimum Principle (PMP)-based conrol, require knowledge of an enire driving cycle and an elevaion profile o resul in opimal performance [4, 5]. Some researchers have shown [1, 6, 7] ha he fuel consumpion of PHEVs is minimized when he ery and engine are used consisenly during an enire rip, such ha he ery sae of charge (SOC) decreases coninuously and reaches he minimum value a he end of he rip. Figure 1 shows he SOC rajecory resuls wih differen conrol sraegies. However, he blended mode conrol ha ensures he minimum fuel consumpion requires more accurae informaion abou he rip, such as driving duraion, driving profile, and so forh. Figure 2: SOC profile and engine ON ime for EV mode and blended mode conrol [6] Our research was conduced under he assumpion ha he vehicle can use he informaion exraced from Inelligen Transporaion Sysems (ITS) over he driving road. This paper describes our invesigaion ino he mehod of using raffic knowledge for predicion of he fuure driving cycle, as well as he opimal conrol sraegy based on PMP, o minimize fuel consumpion. This paper is organized as follows. Secion 2 summarizes he concep of he conrol scheme using he raffic informaion and he navigaion sysem (including a Global Posiioning Sysem [GPS] receiver) in a vehicle. Secion 3 describes he proposed mehod for he driving schedule esimaion. Wih he help of he GPS and ITS, i is possible o direcly predic he fuure driving cycle hrough Dynamic Programming (DP) heory. Secion 4 focuses on he PMP-based conrol and he correlaions beween he prediced driving cycle and he conrol parameer, co-sae p, used in PMP conrol heory. Secion 5 deails he simulaion resuls. The real driving profile from he on-board GPS device and he prediced speed profile are compared, along wih he resuls of he fuel economy improvemen hrough he PMPbased conroller. Finally, Secion 6 presens conclusions abou he uiliy of he proposed conrol mehod. 2 The Conrol Concep of PHEVs Using Traffic Informaion To apply exernal informaion such as he raffic knowledge and he road profile o he opimal conrol for PHEVs, i is necessary o conver various informaion hrough he ransmier of he ITS and GPS ino informaion or daa suiable for he supervisory conrol algorihm. Figure 2 shows he schemaics of he informaion flow among he required modules and a hybrid conrol uni. The Figure 1: Schemaic of he conrol procedure for he PHEV using exernal informaion EVS26 Inernaional Baery, Hybrid and Fuel Cell Elecric Vehicle Symposium 2

3 World Elecric Vehicle Journal Vol. 5 - ISSN WEVA Page 151 hybrid conrol uni, in general, includes he algorihm o deermine he operaing sae of a combusion engine and an elecric driving moor. To calculae he proper oupu power of he componens, he conrol algorihm uses he exernal informaion and operaing sae of he componens, such as he ery s SOC value and signals ha convey informaion concerning he saus of a powerrain. The currenly available ITS and GPS devices provide only limied informaion on raffic circumsances. Therefore, as he iniial par of his sudy, we have developed he ITS model in order o generae he required raffic informaion over each road segmen, such as he locaion and iming of raffic signals, he speed limi, he average speed and rip disance, and so forh. The ITS model uses informaion based on measured daa on cerain real roads o provide he knowledge required for he speed predicion algorihm module. Then he module ha employs he DP algorihm calculaes he predicable speed of a vehicle on he specific segmen of he road, while considering he raffic and road informaion. For he elevaion profile, we use he real daa observed from he GPS device in his sudy, under he assumpion ha he elevaion profile of he overall rip disance can be exraced by he navigaion sysem and 3D-Map [8]. From he esimaed speed profile, he module for he co-sae generaion algorihm can deermine he opimal co-sae o minimize he fuel consumpion, while he final SOC of he ery reaches he lower limi a he end of he rip. Finally, he opimal conrol heory applied o he hybrid conrol uni insananeously compues he raio of he power spli beween he engine and he elecric driving moor by using he prediced co-sae value. 3 Driving Speed Predicion Algorihm The process o generae he speed profile from he raffic informaion is shown in Figure 3. To obain he speed profile for fuure driving, DP heory using a disance-based echnique has been adoped, which reflecs some consrains ha are deermined by raffic informaion on each rip segmen. Generally, because DP is a represenaive echnique o obain opimal policy, in many sudies DP heory has been used o analyze he opimal performance of a hybrid vehicle sysem on a given driving cycle [9 11]. This mehod can be used o minimize he performance index: T J ( x( T)) L(, x( ), u( )) d (1) where, Φ(x(T)) is a penaly funcion o represen consrains on he final SOC, x(t); L( ) is he cos funcion for fuel consumpion; x() is he sae variable ha should be conrolled; and u() is he conrol variable in he sysem. The opimal soluion also should be subjec o he consrains for physical limiaions of componens and he consrains for SOC operaion, as implied by: x f ( x( ), u( ), ) (2) In general, DP requires gridding of he sae and ime variables, and hus he opimal rajecory is calculaed for only discreized values of ime and ery SOC [3]. The DP algorihm explained above, on he oher hand, also can be applied o deermine he fuure speed profile ha saisfies some physical consrains. If he performance measure (or cos funcion) in he DP algorihm could be chosen properly, i is possible o generae he speed profile similar o he driving paern of a real driver. The performance measure for generaing he disance-based speed profile is defined as: J s f s L( x( s), u( s)) ds (3) where, s is posiion, and x(s) and u(s) denoe he sae and conrol variable regarding posiion s, respecively. For he defined disance from he iniial posiion, s, o he final posiion, s f, as a Figure 3: Algorihm process o obain he prediced speed profile on a segmen EVS26 Inernaional Baery, Hybrid and Fuel Cell Elecric Vehicle Symposium 3

4 World Elecric Vehicle Journal Vol. 5 - ISSN WEVA Page 152 segmen of he whole ravel disance, he speed profile ha minimizes he performance index (3) subjec o he equaions of powerrain and speed limi on he road can be obained numerically by solving a DP problem. Here, he cos funcion L( ) a each calculaion sep inuiionally consiss of he driving energy, E drv, he ime consumed in driving, T, and he erms regarding he level of acceleraion, a, and deceleraion, d, as described by: 4 4 L( k) E ( k) w T( k) w a ( k) w d ( k) (4) drv where, k is he sage on he grid of considered disance, and w, w a, and w d are he weighing facors for ime consumed, acceleraion, and deceleraion, respecively. The weighing facors are deermined from map daa found by considering he speed limi, average speed, cruising speed, and lengh of each segmen for raveling disance. Figures 4 and 5 show he predefined value of each weighing facor. Disane [m] Seady Speed [km/h] Figure 4: Weighing facor value for he consumed ime in driving Weighing Facor Weighng Facor for Driving Time Figure 5: Weighing facor value for he level of acceleraion and deceleraion a Weighing Facor for Acc. & Dec Avg Spd/Lim Spd Predicion of he speed profile on a cerain road can be accomplished by selecing only he hree facors regarding driving saes driving ime, acceleraion, and deceleraion on he given disance of he segmen. Figure 6 presens he resul of he shor-range predicion as an example. d The prediced speed profile is similar o he real speed profile, which is exraced from he GPS receiver in a vehicle. We also can observe ha he level of acceleraion and deceleraion of he prediced speed in he figure properly reflecs he driving endency of a real driver. Vehicle speed [km/h] Acceleraion [m/s 2 ] Prediced speed profile Real speed profile Time [sec] Figure 6: Comparison beween prediced speed and real speed for he shor-range predicion The performance of he proposed mehod for he relaively longer disance, including raffic signals, was sudied via measuremen of he required daa on he real road. Figure 7 depics he real driving speed and he informaion for raffic signals on a specific sree near Argonne Naional Laboraory. This informaion is saved in he ITS model. The oal driving disance is 7.5 km, and 13 raffic signals are locaed on he road. Through his informaion, i is possible o esimae he duraion and number of he vehicle s sops. The real and esimaed velociy profiles are shown as a funcion of rip ime in Figure 8. The acceleraion of he real vehicle, unil around 4 seconds, iniially is lower han ha of he prediced speed due o he uphill elevaion of he real road. Excep for he iniial driving sae shown in Figure 8, he speed predicion algorihm can esimae he anicipaed vehicle speed very well during he enire rip. Disance [km] Vehicle Speed & Traffic Signals Driving ime [sec] Figure 7: Traffic signal informaion and comparison of real and esimaed speeds EVS26 Inernaional Baery, Hybrid and Fuel Cell Elecric Vehicle Symposium 4

5 World Elecric Vehicle Journal Vol. 5 - ISSN WEVA Page Real speed profile Prediced speed profile involves only fuel consumpion, he conrol concep minimizes he Hamilonian [3, 5, 7], which is defined as: Vehicle speed [km/h] Aliude [m] Time [sec] Aliude informaion ime [sec] Figure 8: Predicion of he speed profile over a 7.5-km disance and elevaion profile Therefore, if i is realizable o obain he precise informaion from he ITS, GIS, or GPS, hen predicing he vehicle speed on he fixed rip disance can be possible in advance, and he resul is similar o he acual driving paern. The fac ha he fuure speed profile can be obained accuraely is very imporan, since he parameer (i.e., co-sae p) used in he opimal conrol heory should be derived by using he resul of he prediced speed. 4 Opimal Power Managemen Based on PMP Theory An opimal conrol sraegy based on PMP is a promising soluion. I provides a simple soluion for conrolling HEVs or PHEVs and guaranees he bes performance under reasonable condiions [5]. Wih informaion abou he fuure driving schedule and elevaion profile, i is possible o deermine he opimal value for he co-sae, p, depending on he fuure power demand. Then, he PHEV can run in blended mode, so he ery is nearly depleed a he end of he rip. 4.1 Analyical Mehod for he Minimum Fuel Consumpion The opimal conrol based on PMP can be implemened in real-ime applicaions because i is based on insananeous opimizaion. Assuming ha he cos funcion o be opimized H m P ) p( ) SOC ( SOC, P ) (5) fc( where, ṁ fc is he rae of fuel consumpion; p() is an adjusmen variable, which is called co-sae in PMP; and SȮC is a ime derivaive of ery SOC. As saed above, assuming ha minimum fuel consumpion is he goal of he opimal conrol, he problem of PHEVs can be defined as (6) and (7), in which he engine speed, W e, and he engine orque, T e, can be used o deermine he fuel consumpion: min J f L( W, T, ) d (6) subjec o: SOC( ) SOCiniial SOC( f ) SOCdesired SOCmin SOC SOC W min We Wmax Tmin Te Tmax e e max (7) where, L(W e, T e, ) is he rae of fuel consumpion of he engine. The SOC is deermined by a ery model described as: 2 1 V V 4RP SOC Q 2R (8) f ( SOC, We, Te ) Furher T e and W e are resriced by operaing consrains, such as he maximum possible engine speed or he maximum possible engine orque, given by considering he impac of consrains on he componens (i.e., maximum moor speed, maximum orque, or maximum ery power). This opimal conrol problem can be solved wih opimal conrol echniques. When he final ime and he final sae are fixed, he principle requires ha he opimal soluion saisfies he following condiions [7, 12, 13]: H SOC p (9) H p SOC * * * * * H ( SOC, u, p, ) H ( SOC, u, p, ) Equaion (9) is he necessary condiion of he opimal problem by PMP heory. I is necessary o find he opimal conrol, u, which saisfies (9). EVS26 Inernaional Baery, Hybrid and Fuel Cell Elecric Vehicle Symposium 5

6 World Elecric Vehicle Journal Vol. 5 - ISSN WEVA Page Esimaing he Iniial Co-sae from he Prediced Speed Profile The opimal value of he co-sae p() should ensure ha he final SOC (SOC final ) is equal o he desired SOC (SOC desired ) a he end of he enire rip. The opimal co-sae value ha is subjec o he above condiion on a given driving schedule can be found by running he simulaion repeaedly on he various iniial values of p( ) or he shooing mehod [1, 14]. However, for he opimal conrol algorihm o have he poenial for a real-ime implemenaion, i needs o esimae he value of he co-sae close o he opimal value. If he fuure driving cycle is known a priori hrough he mehod proposed in Secion 3, i is possible o predic he iniial cosae value p( ) of he differenial equaion ṗ in (9), which can execue he opimal conrol o minimize he fuel consumpion on he esimaed rip schedule. In his sudy, p( ) can be formulaed as (1) and (11) because he iniial co-sae has a significan influence on he effecive SOC drop rae, SȮC drv,eff, and he usable ery energy, E [MJ], a he iniial driving sae: p f ( SOC a( E drv, eff, E ) SOC ) drv, eff b( E ) (1) n1 n1i a ( E) Cai E ( n 4) (11) i1 n1 n1i b( E) Cbi E ( n 4) i1 where, a and b are he coefficiens deermined by he useable ery energy, and C ai and C bi are he consan values, as shown in Table 1. Table 1: Numerical values of he coefficiens in equaion (11) C a1 C a2 C a3 C a4 C a , , ,126 C b1 C b2 C b3 C b4 C b The effecive SOC drop rae, SȮC drv,eff, in equaion (1), which should be calculaed from he speed profile, can be wrien as: SOC drv, eff drv, eff (12) Tdrv, eff SOC where, ΔSOC drv,eff denoes he decreased value of SOC when considering only he ery s discharging sae during he enire rip, and ΔT drv,eff is he effecive driving ime, excep for he ime during he sop and deceleraion condiion of a vehicle. For insance, his relaion is shown in Figure 9. The effecive SOC profile represens he cumulaive value of posiive or negaive deviaion of he SOC wih respec o ime, which is divided ino he propulsive driving and he regeneraive driving condiions. Figure 9: Effecive SOC profile divided ino he propulsion and he regeneraion sae Because we canno direcly observe ΔSOC drv,eff if we do no execue he simulaion on a compuer, alernaively one can use a cumulaive deviaion of SOC by regeneraion, ΔSOC reg,eff, o predic ΔSOC drv,eff as follows: (13) SOC drv, eff SOCi SOC f SOCreg, eff where, ΔSOC reg,eff is he increased value of SOC when considering only he ery s charging hrough he regeneraive breaking. This value can be direcly deermined if he upcoming driving paern is prediced as shown in Figure 1(a). Vehicle speed [km/h] Time [sec] (a) Real speed profile and esimaed speed profile Power [kw] Required wheel power Real speed profile (b) Required posiive and negaive power for he driving speed Figure 1: Speed profile and required power Prediced speed profile Time [sec] EVS26 Inernaional Baery, Hybrid and Fuel Cell Elecric Vehicle Symposium 6

7 World Elecric Vehicle Journal Vol. 5 - ISSN WEVA Page 155 Nex, he fuure power demand wih respec o driving ime is deermined as described by: P ( ) F ( ) v( ) (14) where, P () is he required racive power, F () is he racive force, and v() is he vehicle speed. The racive force F is deermined by he longiudinal vehicle dynamics model [15]. Figure 1(b) shows he posiive required power for propulsion and he negaive required power for deceleraion. Then we can surmise he increasing amoun of ery SOC, ΔSOC reg,eff, from regeneraive breaking power, (P ()<), over he enire rip ime by using equaions (8), (15) and (16), under he assumpion ha he volage and resisance of he ery are of average value: SOC reg T eff SOC, ( P( )) d (15) P( ) reg P ( ) ( P ( ) ) (16) where, P () is he elecrical power charged o he ery, and η reg is he raio of elecrical braking wih respec o he oal braking power a a wheel. Figure 11 shows he resuls of he opimal SOC effecive drop o he various driving disances, which is calculaed hrough he backwardlooking simulaion on repeaed Urban Dynamomeer Driving Schedule (UDDS) cycles. Effecive SOC SOC effecive for regen. SOC effecive for prop Effecive ime [sec] Figure 11: Resuls of he effecive SOC drop using he backward-looking simulaion based on he opimal conrol heory Table 2 deails he validaion resuls of he prediced co-sae and he opimal value calculaed from driving cycles repeaed five imes, as well as he various ery capaciies. In a PHEV using a 3.4-kWh ery pack, he vehicle can be operaed in elecric vehicle (EV) mode (charge-depleing [CD] mode) only on a single driving of a UDDS cycle. Thus, he predicion of he iniial co-sae on jus one cycle could be meaningless, since no using he engine is more efficien during he enire rip. When he raveling disance is known o be less han or equal o he vehicle s elecric range, he powerrain can run in is all-elecric mode [1]. This is also he same as he PHEV using 5.2-kWh ery capaciy, if he driving cycle is repeaed wice. Excep for he rip disance on which he PHEV can be operaed in EV mode only, he co-sae esimaed from (1) is very close o he opimal co-sae value. Table 2: Resuls of he co-sae validaion on he UDDS cycle Toal Baery Energy 1.4 kwh 3.4 kwh 5.2 kwh Co-sae UDDS 1 UDDS 2 UDDS 3 UDDS 4 UDDS 5 Opimal Prediced Error.48%.36%.8% -.6% -.12% Opimal Prediced EV mode Error.8%.9%.23% -.5% Opimal Prediced EV EV mode mode Error.24%.19%.19% 5 Simulaion Resuls The performance of he PMP-based conrol, wih he speed profile predicion described in Secion 3, is sudied by comparing he conrol mehod using a ypical power managemen scheme for PHEVs. This scheme is o run he PHEV in is all-elecric mode unil he ery is nearly depleed and swiched o a charge-susaining (CS) mode, and hen o run he PHEV similar o an HEV [1]. 5.1 Vehicle Model The vehicle model used in his sudy is a powerspli hybrid sysem ha has a single planeary gear se as a power-spli device, like ha shown in Figure 12. All daa for he componen models and vehicle model (shown in Table 3) are based on a 24 Toyoa Prius in Auonomie, a sofware ool developed by Argonne Naional Laboraory [7, 16]. The oal ery energy was decreased o 3.4 kwh o verify he opimaliy for he esimaed driving condiion. Moor 2 Engine Ring gear Carrier Sun gear Final drive Moor 1 Vehicle Figure 12: Power-spli hybrid sysem used in his sudy EVS26 Inernaional Baery, Hybrid and Fuel Cell Elecric Vehicle Symposium 7

8 World Elecric Vehicle Journal Vol. 5 - ISSN WEVA Page 156 Table 3: Vehicle parameers used in he simulaion Vehicle mass 1,49 kg Engine 1,5 cc Moor 1 25 kw (peak power: 5 kw) Moor 2 15 kw (peak power: 3 kw) Baery 3.4 kwh Planeary 2.6 (78/3) gear raio Final gear raio Rolling resisance vehicle speed Fronal area 2.25 m 2 Drag coefficien.29 Wheel radius.35 m Air densiy 1.23 kg m Evaluaion and Resuls To evaluae he performance of he proposed conrol mehod, we use he real speed daa saved in he GPS device afer driving a vehicle over a disance of almos 15 km. The speed predicion is carried ou hrough he raffic informaion on he road. Then he final speed profile for he simulaion is repeaed wo imes for long disance, as shown in Figure 1(a). In his sudy, he simulaion uses he iniial SOC, SOC i, as 8% and allows he sysem o consume he elecrical energy unil he final SOC, SOC f, falls o 3%. The fuel consumpion analysis requires he following procedure. Firs of all, he fuure speed profile is prediced hrough he raffic informaion on he segmens of he real road. Nex, we calculae he iniial co-sae from he prediced speed profile. Finally, he PMP-based conroller using he co-sae is applied o he vehicle model driving on he real speed profile. The value of he esimaed co-sae and oher regarding values for he prediced speed profile are shown in Table 4. The esimaed P is very close o he opimal value P * obained from he backward-looking simulaion. Figure 13 shows he fuel consumpion resuls on he real speed profile wih respec o he conrol mehods, he CD+CS mode conrol and he PMP-based conrol. SOC drv, eff Table 4: Resuls of he co-sae predicion T drv, eff 1,416 seconds S OC drv, eff Esimaed co-sae p Opimal co-sae p * Error.315% In he PMP mode conrol, he fuel usage coninuously increases wih driving ime as blended mode conrol, while he decreasing rae of he ery SOC value on almos he enire rip is lower han ha of he CD+CS mode conrol, as shown in Figure 14. The final SOC of he PMPbased conrol does no exacly reach he minimum value a he end of he rip. This occurs because he simulaion is execued over he real speed profile afer we derive he iniial co-sae from he esimaed speed profile. Noneheless, as a consequence, he PMP-based conrol resuls in smaller fuel consumpion a he end of he given rip, and he fuel economy is increased by around 17% compared wih he CD+CS mode conrol mehod. Table 5 summarizes hese findings. The oupu power of he engine for each conrol mehod is shown in Figure 15. Table 5: Comparison of he fuel consumpion Conrol Mehod for PHEVs Fuel Economy (Km/L) Final SOC FE Increasing rae CD+CS conrol Ref. Predicion-based PMP conrol Fuel mass [g] % Comparison of fuel consumpion CD+CS mode conrol PMP based conrol ime [sec] Figure 13: Fuel consumpion resuls for he conrol mehod SOC Comparison of SOC rajecory CD+CS mode conrol PMP based conrol ime [sec] Figure 14: SOC rajecory for he conrol mehod Power [kw] CD+CS mode conrol PMP based conrol Engine oupu power ime [sec] Figure 15: Engine oupu power for he conrol mehod EVS26 Inernaional Baery, Hybrid and Fuel Cell Elecric Vehicle Symposium 8

9 World Elecric Vehicle Journal Vol. 5 - ISSN WEVA Page Conclusion This sudy has invesigaed he opimal conrol sraegy, PMP-based conrol, for PHEVs by using predicion of he fuure driving schedule. To predic he fuure driving paerns, if he raffic informaion is available, he DP mehod based on driving disance can be used over he given rip. This is achieved by minimizing he proposed cos funcion on each segmen. The resul of he generaed speed profile can properly esimae he driving paern of he driver. Deriving he co-sae used in he opimal conrol from he prediced speed profile is a very imporan procedure o minimize he fuel consumpion over he enirey of he ravel. The co-sae generaion algorihm is applied in order o deermine he parameers wih respec o he required power deduced from he prediced driving cycle. The proposed co-sae generaion model can esimae an iniial co-sae similar o he opimal co-sae. Using he parameers, he PMP-based conrol algorihm insananeously calculaes he opimal power-spli raio of power sources. Simulaion resuls indicae ha his approach guaranees he minimizaion of fuel consumpion on he rip disance beween he origin and desinaion under reasonable assumpions. Acknowledgmens This work was suppored by DOE s Vehicle Technology Office under he direcion of Lee Slezak and he Naional Research Foundaion of Korea (NRF) gran funded by he Korea governmen (MEST) (No ). The submied manuscrip has been creaed by UChicago Argonne, LLC, Operaor of Argonne Naional Laboraory ( Argonne ). Argonne, a U.S. Deparmen of Energy Office of Science laboraory, is operaed under Conrac No. DE- AC2-6CH The U.S. Governmen reains for iself, and ohers acing on is behalf, a paidup nonexclusive, irrevocable worldwide license in said aricle o reproduce, prepare derivaive works, disribue copies o he public, and perform publicly and display publicly, by or on behalf of he Governmen. References [1] C. Zhang, e al., Roue Preview in Energy Managemen of Plug-in Hybrid Vehicles, IEEE Trans. Con. Sys. Tech., ISSN , 2 (212), [2] H. Banvai, e al., A Rule Based Energy Managemen Sraegy for Plug- in Hybrid Elecric Vehicle (PHEV), American Conrol Conference, ISSN , 29, [3] A. Sciarrea, e al., Conrol of Hybrid Eelecric Vehicles, IEEE Conrol Sysems, ISSN X, 27 (27), 6-7. [4] Q. Gong, e al., Trip Based Opimal Power Managemen of Plug-in Hybrid Elecric Vehicles, IEEE Trans. on Veh. Tech., ISSN , 57 (28), [5] N. Kim, e al., Opimal Conrol for Hybrid Elecric Vehicles Based on Ponryagin s Minimum Principle, IEEE Trans. Con. Sys. Tech., ISSN , 19 (211), [6] V. Marano, e al., Comparaive Sudy of Differen Conrol Sraegies for Plug-in Hybrid Elecric Vehicles, SAE conference, , 29. [7] N. Kim, e al., A Jump Condiion of PMP- Based Conrol for PHEVs, Journal of Power Sources, ISSN , 196 (211), [8] K. Moran, e al., Digial Maps, Conneciviy and Elecric Vehicles Enhancing he EV/PHEV Ownership Experience, SAE In. Journal, Passeng. Cars-Elecron. Elecr. Sys, 3 (21), [9] C. Musardo, e al., A-ECMS: An Adapive Algorihm for Hybrid Elecric Vehicle Energy Managemen, in Proc. 44h IEEE Conf. Decision Conrol, 25 European Conrol Conf., 25, [1] Y. Zhu, e al., A Four-Sep Mehod o Design an Eenergy Managemen Sraegy for Hybrid Vehicles, in Proc. 24 American Conrol Conf., 24, [11] C.-C. Lin, e al., Power Managemen Sraegy for a Parallel Hybrid Elecric Truck, IEEE Trans. Con. Sys. Tech., ISSN , 11(23), [12] P. Pisu, e al., A Comparaive Sudy of Supervisory Conrol Sraegies for Hybrid Elecric Vehicles, IEEE Trans. Con. Sys. Tech., ISSN , 15 (27), [13] J. Jeong, e al., Fuel Economy Analysis of a Parallel Hybrid Bus Using he Opimal Conrol Theory, VPPC211, 211. [14] N. Kim, e al., Opimal Conrol of a Plug-in Hybrid Elecric Vehicle (PHEV) Based on Driving Paerns, EVS24, 29. EVS26 Inernaional Baery, Hybrid and Fuel Cell Elecric Vehicle Symposium 9

10 World Elecric Vehicle Journal Vol. 5 - ISSN WEVA Page 158 [15] T. Keulen, e al., An Adapive Sub-Opimal Energy Managemen Sraegy for Hybrid Drive-Trains, 28 IFAC, 28, [16] A. Rousseau, e al., Inegraing Daa, Performing Qualiy Assurance, and Validaing he Vehicle Model for he 24 Prius Using PSAT, SAE Technical Paper, , 26. Auhors Daeheung Lee Ph. D. Candidae, Seoul Naional Universiy, Souh Korea ideal2@snu.ac.kr Daeheung Lee received his bachelor s degree in Deparmen of Mechanical Engineering from Hanyang Universiy, Souh Korea, in 26. He is currenly working owards Ph.D. degree in School of Mechanical and Aerospace Engineering a he Seoul Naional Universiy. Suk Won Cha Associae Professor, Seoul Naional Universiy, Souh Korea swcha@snu.ac.kr Suk Won Cha received his M.S. and he Ph.D. degree in Deparmen of Mechanical Engineering from Sanford Universiy, in 1999 and 24, respecively. His research ineress are fuel cell sysems, design of hybrid vehicle sysems and applicaion of nanoechnology o energy conversion devices. Aymeric Rousseau Program Manager, Argonne Naional Laboraory, USA arousseau@anl.gov Aymeric Rousseau received his Maser of Science in Indusrial Sysems from EIGSI in La Rochelle, France, in He is currenly leading Argonne s vehicle modeling and simulaion group. Namwook Kim Posdocoral Appoinee, Argonne Naional Laboraory, USA nakim@anl.gov Namwook Kim received his bachelor s degree and docor s degree from Seoul Naional Universiy, Souh Korea. He is currenly working on modeling dynamic sysems and developing supervisory conrol sraegies for Auonomie. EVS26 Inernaional Baery, Hybrid and Fuel Cell Elecric Vehicle Symposium 1