Hierarchical Charging Management Strategy of Plug-in Hybrid Electric Vehicles to Provide Regulation Service

Transcription

1 3rd IEEE PES Innovative Smart Grid Technologie Europe (ISGT Europe), Berlin Hierarchical Charging Management Strategy of Plug-in Hybrid Electric Vehicle to Provide Regulation Service Shouxiang Wang, Senior Member, IEE E, Liang Han, Dan Wang, Member, IEEE, Mohammad Shahidehpour, Fellow, IEEE, Zuyi Li, Senior Member, IEEE Abtract Plug-in Hybrid Electric Vehicle () can be treated a mobile ditributed energy torage unit in a power grid. Thi paper addree the feaibility of achieving ytem ancillary ervice with large-cale integration. The charging model of ingle and the tatitic power demand model for charging proce of large-cale are preented. Baed on thee model, a hierarchical large-cale charging management trategy with demand repone i propoed. The propoed trategy not only meet the requirement of ytem dipatching but alo i cutomer atifaction contrained. Several cae are tudied to verify the influence of cutomer atifaction degree and tarting time of charging management control on the potential of to provide the auxiliary ervice adjutment. Simulation reult alo how that the number of involved in the charging management proce, which i motly influenced by the etting of cutomer atifaction degree and the moment when plug in the grid, ha a great influence on the performance of charging management trategy. Index Term -- Plug-in Hybrid Electric Vehicle (), ancillary ervice, charging model, charging management trategy I. INTRODUCTION CCORDING to development planning of new energy Aautomotive indutry (-) in China, the market penetration of electric vehicle (EV) and plug-in hybrid electric vehicle () will be 5 million by. The total aggregate energy torage capacity will be about million kw, auming the average charge held by each i 4kW. It i clear that, with large cale uptake of electric vehicle, the total charging power will be enormou []. Without any intelligent charging algorithm, the large-cale addition of into the power grid will increae the peak load, which require expenive grid improvement and more peak load power plant, reulting in increaed running cot and operating demand for generation, tranmiion and ditribution [][3][4][5][6]. Several EV battery charging model are propoed in [7] uing tatitical method. In thi literature, the impact of four charging model on the maximum load in ditribution network are tudied where, for example, the maximum grid load will increae by 7.9% with a % EV market penetration rate, while a % level of EV S. Wang and L. Han are with Key Laboratory of Smart Grid of Minitry of Education, the School of Electrical Engineering and Automation, Tianjin Univerity, Tianjin 37, China ( xwang@tju.edu.cn, hanliang@63.com). Dan Wang i with the Tianjin electric power company, Tianjin, 3, China. M. Shahidehpour and Z. Li are with the Electrical and Computer Engineering Department, Illinoi Intitute of Technology, Chicago, IL 666 USA ( m@iit.edu, lizu@iit.edu). Thi work wa upported in part by National Natural Science Foundation of China (NSFC 5837, 57798) and National Baic Reearch Program of China (9CB978). penetration will lead to peak load increae of 35.8%. In [8], the daily power demand curve of an EV i hown uing a Monte Carlo method. A urvey on the actual load in Beijing City and Shanghai City in China how that a large number of electric vehicle adding into the grid will increae peak load if there i no charging trategy. The batterie ued by are different from other type of load. They can be conidered a controllable load in a power grid becaue they may be curtailed for ignificant period of time (e.g., everal hour) without impact on end-ue function [9]. The battery can be fully ued by etablihing a reaonable load control trategy. When a i connected to the grid, the charging tate of a battery can be determined by dipatching the battery information back to the ytem. A a reult, the will repone to the electricity price ignal and excitation mechanim [][]. Much reearch ha been done on load control which mainly focue on uing controllable thermodynamic device, uch a water-heater and air conditioner, to control load [][3][4]. In [5], a tate-queuing model i put forwarded, within which thermodynamic load can dynamically repond to a price ignal. With centralized management and control, thermodynamic load can be regarded a a virtual torage device and the load will repond to the output volatility of wind-turbine and other new energy power plant [6]. Controlling load with can not only olve the peak charging load demand caued by large cale adoption, but alo can upply tored energy in, which can be treated a a mobile ditributed energy torage unit in the grid, a regulation, load hifting, emergency ecurity, pinning reerve and other ancillary ervice uch a energy torage and frequency regulation [7][8].The traditional ytem control and load control method are compared in [9]. In addition, a load control trategy framework i decribed and a dual objective control method and the communication framework for controlling are propoed a well. Some paper, relying on pricing trategie, have been aimed at hifting the charging load to off-peak hour [][]. Beide charging control trategy of, reearche alo tudying the impact of providing Vehicle-to- Grid(VG) [8][], thee literature propoe the VG charging and dicharging management trategy, and include the calculation of battery current available capacity, load hifting method and battery contraint. Electric vehicle can be regarded a a ort of ditributed energy torage unit which can let renewable energy ource better integrate into power grid [3][4][5]. VG will improve grid flexibility, reliability and energy efficiency and other benefit uch a lowing down the upgrade contruction requirement for generation, tranmiion, ditribution //$3. IEEE

2 Thi paper addree the feaibility of achieving ytem ancillary ervice with. Firt, the charging model of a, including the main obervation and control parameter i decribed. Second, a charging management trategy i propoed for large-cale to achieve ancillary ervice of regulation baed on the above model. Finally, it point out the factor on the performance of ytem control with : uer electricity upply atifaction etting and the moment when plug in the grid. II. CHARGING MODEL FOR SYSTEM ANCILLARY SERVICES A. Charging proce and power demand model of ingle Currently, a mainly ue lithium battery. Comparing to the whole proce, the tarting and ending of battery charging proce are relatively hort and can be ignored. A implified model, a hown in Fig., replace the actual charging proce and the power load of a battery can be conidered a contant. Power conumption (kw) Actual charging power Simplified charging power State of Charge Charging time Fig.. The actual and the implified proce of charging for State of charge (SOC) i the ratio of current battery capacity to it full charge capacity, commonly expreed in percentage. The range i % to % (% mean empty while % mean full). The relationhip between SOC and charging power during the charging proce i hown in Fig. [7]. If the battery power i P c and the ampling period i minute, the linear relationhip between battery SOC and time i determined by (): Or in recurrence form: SOC (%) Pc SOC( t) = t % () 6E SOC( t +Δ t) = SOC( t) + Δ t % () 6E Where t i the moment when a i plugged in the grid ; P c i charging power (kw), Δt = min i the ampling period, E i the battery capacity (kwh). P c B. Aumption of large-cale charging proce and quantitative aement Thi paper addree the demand power model for charging proce of large-cale and achieving the relevant control of demand repone. In general, the load power for large-cale charging i largely determined by three factor: the moment being plugged in the grid, initial SOC at the beginning of charging and the quantity of. For the firt two factor, the aumption for the charging proce of large-cale are a follow: Each can tart charging immediately once being plugged in the grid by ignoring the tart-up delay of charging; All are charged at the rated power; Initial SOC of each i ; Each will be charged to full capacity; The initial moment when are plugged in the grid meet the normal ditribution and the probability denity function i (3) [8]: ( τ μ ) exp ( ) 4 μ < τ σ π σ T () τ = (3) ( τ+ 4 μ ) exp τ ( μ ) < σ σ π where: μ =7.6, σ = 3.4 Take 6: AM a the reference tarting time. According to (3), the probability ditribution of charging at the tarting time i hown in Fig. with the number of imulation being et to,. The tarting time of charging can be een a the returning time of the lat travel. Therefore, it i determined by the uer travel habit. In general, at 7:-8: PM, mot owner return home, and many are plugged in the grid for charging, reulting the correponding peak load. Probability denity Charging time Fig.. The probability denity of charging time The effectivene of regulating auxiliary ervice depend on the adequate quantity of [5]. We imulate the daily average charging power on different cale of according to () and (3), curve are hown in Fig.3. The daily average charging power i determined by (4):

3 3 P ave = N N i= where: P ci i charging power of the i-th (kw); N i the quantity of. In Fig.3, four load curve are obtained when the number of i et to,,, and, repectively. When N i,, or, the imulated curve are fluctuant and have no regularity, ince the quantity of i not ufficient to provide effective regulatory capacity. Therefore, the cae of N = i ued for numerical imulation to dicu controllability of..5.5 P ci : = : = : =, : =, (4) Charging management center Power Grid Management Charging management center ON/ OFF Unit Control ignal Charging management center3 n n n.5 Central Controller Sytem regulation target Power Signal 5 5 time Fig.3. load profile in different cale III. DEMAND RESPONSE ORIENTED CONTROL STRATEGY OF CHARGING MANAGEMENT BASED ON A. The framework of charging management In thi paper, a hierarchical management framework i propoed a hown in Fig.4. The power grid i divided into different area according to the location (e.g. reidential quarter, office building, etc.). In a certain geographic area, a charging management center i et up for the management of the in the area and provide communication interface between the and the dipatching center. The information of all i ent to the charging management center and ued to build up a group-reponding model, in order to decribe the repone ability of thi group to the ytem control ignal. According to the ytem control demand, the dipatching center end control ignal to the charging management center to increae (or decreae) the charging load. Baed on the charging management trategy in the next ection, the charging management center will decide which hould be charged (or top charging), and end the final information about the power demand back to the dipatching center to provide a reference for the next control. Fig.4 The framework of hierarchical charging management B. Demand repone oriented control trategy of charging management Once a i under control, it battery charging tatu would be managed by the dipatching center. The charging proce i hown in Fig. 5, and () can be rewritten a: Pc SOC( t +Δ t) = SOC( t) + nc Δ t % (5) 6E Where: n c i charging tatu at time t ( n c = mean being charged while n c = mean not being charged). ct i the charging cumulative time; nct repreent the no charging cumulative time. ct and nct can be recorded by the mart meter or other advanced meaurement ytem. SOC (%) SOC Charging tatu nct Charging time Fig.5 The Controlled proce of charging When ytem ancillary ervice are achieved by load control, both the requirement of ytem regulation and the ct Charging tatu

4 4 uer atifaction with electricity upply hould be met [6][9]. The atifaction of uer mean to finih charging within the pecified time and enure the normal ervice life of the battery. Aume the target power of grid at t i P trg (t), the total power conumed by all i P tot (t), take P trg (t)>p tot (t) for example to introduce how to determine the charging tate of a certain at t. Firtly, the charging management center collect the information of nct at t ; then, according to the nct and the remaining time, the center ort the ; if there are plenty of controllable, the will be equential changed to not charging until the error of P trg (t)-p tot (t) i acceptable, otherwie all are changed into not charging. IV. CASE STUDY A. Cae decription According to the data of Chevrolet' electric vehicle Volt, the charging power i ditributed evenly within 3-4 kw with battery capacity 6 kwh. A introduced in Section II.B, the number of i et to,. The imulation duration i from 6: AM to 6: AM the next day (t = -44), and the imulation tep i min. B. The impact of different AT value on the ytem regulation Taking AT a 3 min, 6 min and min, the impact of different value of AT on the ytem regulation i tudied. The imulation reult are how in Fig.6 to Fig.9. A whole day i divided into three period to reearch the relative average error for a number of controlled with different value of AT. The reult are hown in Table I. The upper and lower boundarie of load power are recorded a well. Relative average error (RAE) i determined by: e = k k t = P trg ( t) P P ( t) trg tot ( t) % There are two kind of the number of controlled, one i N, the average number of which can be changed to no charging tate per minute, the other i N aying the average number of which can be changed to charging tate per minute. The dipatching center can ue the upper and lower boundarie of load power a a reference for ytem control. In thi way, the aggregated load of can be taken a a virtual power generator. The upper boundary i the actual power conumption added by the power of which can be changed to charging tate per minute; the lower boundary i the total power conumption of which mut be at charging tate. Fig.7 i the zoom area for the rectangle area in Fig.6. Different time period have different regulating error with fixed AT. At the beginning (-48 min), limited number of plugged in the grid lead to the limited controllable which reult in fully failing to repond to dipatch; In the mid-term (48-96 min), with more plug in grid, controllable gradually increae while the regulating (6) error reduce. In the latter of regulation (96-44 min), mot finih charging and exit grid. The number of controlled gradually reduce while the regulating error increae with ytem control capacity reducing Target Controlled Uncontrolled Lower Boundary Upper Boundary 5 5 time Fig.6. Simulated power variation in regulation operation(at=3 min, controlled duration i min to 44 min) Fig.7 The drawing of partial enlargement Fig.8. Simulated power variation in regulation operation(at=6 min, controlled duration i min to 44 min)

5 Target Lower Boundary Controlled Upper Boundary Uncontrolled 5 5 time Fig.9. Simulated power variation in regulation operation(at= min, controlled duration i min to 44 min) TABLE I COMPARISON OF SIMULATION RESULTS WITH DIFFERENT AT VALUES AT 3 6 Period Controllable amount N N RAE(%) N 3 7 N RAE(%) N N RAE(%) be plugged in the grid and enough charging time hould be accumulated, both of which limit the controllable capacity. Since ytem will accumulate controllable capacity over time, a to the ituation where the target value i greater than the actual value, the performance of control in the econd period (t = 7-44 min) i better than the firt one (t = -7 min). C. The impact of different tart time on the ytem regulating When AT i 3 min, et the tart time t to 7 min, and tudy the impact of different tart time on the actual ytem regulating. Fig. how the imulation reult. Table II compare the imulation reult with different controlled duration and indicate that, if the tarting time delay, the number of controllable at each period will be reduced, and the ytem available controllable capacity will be retricted with increaing error. Simulation reult demontrate, with the fixed AT value, the adjutable capacity of ytem at the ame moment i relevant to the duration of control. The longer the controlling lat, the larger the adjutable capacity i. Relative average error -RAE(%) Fig.. Simulated power variation in regulation operation (AT=3 min, controlled duration i 7 min to 44 min) TABLE II COMPARISON OF SIMULATION RESULTS WITH DIFFERENT CONTROLLED DURATIONS( AT=3MIN) Starting time 7 Period N Controllable amount -44 N RAE(%) N N RAE(%) Fig.. Comparion of ytem regulating error with different AT value A AT increae, uer requirement of electricity upply atifaction increae and the number of, which can be ued by ytem control, reduce. A a reult, the error between the target power and the actual power will increae, then, the average relative error increae in the ame tage. Judging from Fig.6 - Fig.9, the ytem control error will be great when the target value i greater than the actual one. Thi i becaue, to achieve controlling, ufficient capacity hould V. CONCLUSIONS Thi paper addree a framework uing load to participate in the ytem ancillary ervice. The main challenge lie in how to meet ytem regulating requirement while enuring uer atifaction of electricity upply including enuring battery life, finihing charging within the pecified time. The B. Demand repone oriented charging management trategy propoed in thi paper provide a mean to achieve thee goal.

6 6 The performance of charging control trategy i determined by the number of, while cae tudie how that the AT and tart time will influence the number of controlled on the ytem regulation. Thi paper only addree demand-repone technology of charging proce. The following iue need to be further tudied in the further tudy. The comprehenive controlling of PEHV with ditributed generation. Adopting VG (Vehicle-to-Grid) technology, uing a equivalent energy torage device to provide electricity when ditributed generation top working or pare capacity i inufficient, and often the impact of ditributed generation on grid intermittent. VI. REFERENCE [] Luther Dow, Mike Marhall, Le Xu, Julio Romero Agüero, H. Lee Willi, A Novel Approach for Evaluating the Impact of Electric Vehicle on the Power Ditribution Sytem, in IEEE Power & Energy Society General Meeting, pp.-6. [] Liu, Ryan, Dow, Luther; Liu, Edwin. A urvey of PEV impact on electric utilitie, in IEEE PES Innovative Smart Grid Technologie Conference Europe, pp.-8. [3] K. Clement-Nyn, E. Haeen, and J. Drieen, The impact of charging plug-in hybrid electric vehicle on a reidential ditribution grid, IEEE Tran. Power Syt., vol. 5, no., pp ,.. [4] L. Pieltain Fernandez, T. Gomez San Roman, R. Coent, C. M. Domingo, and P. Fria, Aement of the impact of plug-in electric vehicle on ditribution network, IEEE Tran. Power Syt., vol. 6, no., pp. 6-3, Feb.. [5] K. Dyke, N. Schofield, and M. Barne, The impact of tranport electrification on electrical network, IEEE Tran. Ind. Electron., vol. 57, no., pp , Dec.. [6] J. Waddell, M. Rylander, A. Maitra, J. A. Taylor, Impact of plug in electric vehicle on Manitoba Hydro' ditribution ytem, in Electrical Power and Energy Conference (EPEC), pp [7] Kejun Qian, Chengke Zhou, Malcolm Allan, Yue Yuan. Modeling of Load Demand Due to EV Battery Charging in Ditribution Sytem, IEEE Tran. Power Syt., vol. 6, pp.8-8, May. [8] Tian Liting, Shi Shuanglong, Jia Zhuo. A Statitical Model for Charging Power Demand of Electric Vehicle, Power Sytem Technology, vol. 4, pp.6-3, March. [9] D. Callaway and I. Hiken, Achieving controllability of electric load, in Proc. of IEEE, vol. 99, pp [] Aline Senart, Scott Kurth, Gaëlle Le Roux,Aement, Framework of Plug-in Electric Vehicle Strategie, in t IEEE International Conference on Smart Grid Communication (Smart Grid Comm.), pp [] E. Sortomme, K. W. Cheung, Intelligent Dipatch of Electric Vehicle Performing Vehicle-to-Grid Regulation, in IEEE International Electric Vehicle Conference (IEVC), Greenville, SC, USA,. [] Jean-Charle Lament, Guy Deaulnier, Roland P. MalhamC, FranGoi Soumi, A column generation method for optimal load management via control of electric water heater, IEEE Tran. Power Syt. Vol., pp , Augut 995. [3] H. Salehfar, P. J. Noll,Fuzzy, B.J. LaMere, M.H. Nehrir, and V.Gerez, Fuzzy Logic-baed direct load control of reidential electric water heater and air conditioner recognizing cutomer preference in a deregulated environment.in IEEE Power Engineering Society Summer Meeting, 999, pp [4] Khalid Elgazzar, Howard Li, Liuchen Chang, A centralized fuzzy controller for aggregated control of dometic water heater, in 9 Electrical and Computer Engineering, pp [5] N. Lu and D. P. Chain, A tate queuing model of thermotatically controlled appliance, IEEE Tran. Power Syt., vol. 9, no. 3, pp , Aug. 4. [6] D. Callaway. Tapping the energy torage potential in electric load to deliver load following and regulation, with application to wind energy, Energy Converion and Management, vol.5, pp.389 4, May 9. [7] S. Han and K. Sezaki, Development of an optimal vehicle-to-grid aggregator for frequency regulation, IEEE Tran. Smart Grid, vol., no., pp. 65 7, Jun.. [8] E. Sortomme and M. A. El-Sharkawi, Optimal charging trategie for unidirectional vehicle-to-grid, IEEE Tran. Smart Grid, vol., no., pp. 3 38, Mar... [9] Ian Hiken, Duncan Callaway. Achieving Controllability of Plug-in Electric Vehicle, in 9 Vehicle Power and Propulion Conference, pp. 5. [] M. Galu and G. Anderon, Demand management of gridconnected plug-in hybrid electric vehicle (phev), in 8 IEEE Conference in Energy 3, pp. 8. [] Samadi, P., Mohenian-Rad, A., Schober, R., Wong, V.W.S., Jatkevich, J., Optimal real-time pricing algorithm baed on utility maximization for mart grid, in Proc. of the t IEEE International Smart Grid Communication Conference, pp [] W. Kempton, J. Tomi c, Vehicle-to-grid power fundamental: calculating capacity and net revenue, J. Power Source 44, pp , April, 5. [3] W. Kempton, J. Tomi c, Vehicle-to-grid power implementation From tabilizing the grid to upporting large cale renewable energy, J. Power Source 44 pp. 8 94, April, 5. [4] H. Lund, W. Kempton. Integration of renewable energy into the tranport and electricity ector through VG Energy Policy, vol 36, no. 9, pp , 8. [5] Junji Kondoh, Ning Lu, Donald J. Hammertrom, "An Evaluation of the Water Heater Load Potential for Providing Regulation Service," in Proc. IEEE Power and Energy Society General Meeting, pp.-8 VII. BIOGRAPHIES Shouxiang Wang(SM ) received the B.S. and M.S. degree from Shandong Univerity of Technology, Jinan, China, in 995 and 998 repectively and Ph.D. degree from Tianjin Univerity, Tianjn, China, in, all in electrical engineering He i currently a Profeor in the School of Electrical Engineering and Automation at Tianjin Univerity, China. Hi main reearch interet are ditributed generation, microgrid and mart ditribution ytem. Liang Han received the B.S. degree in electrical engineering from Jilin Univerity, Changchun, China, in 9. He i currently puruing the PhD degree in the School of Electrical Engineering and Automation at Tianjin Univerity, China. Hi reearch field i ditributed generation and mart grid. Dan Wang (M ) received the B.S. and M.S. degree from Hohai Univerity, Nanjing, China, in 3 and 6 repectively and Ph.D. degree from Tianjin Univerity, Tianjin, China, in 9, all in electrical engineering. A a PICS pot-doctoral fellow at the Univerity of Victoria, Victoria, BC Canada from to, hi reearch focu on evaluating technical, economic and environmental feaibility of renewable energy integration, demand-ide management and power ytem ancillary ervice analyi. He i currently working in Tianjin Electric Power Company, Tianjin, China. Mohammad Shahidehpour (F ) i Bodine Profeor and Chairman in the Electrical and Computer Engineering Department at Illinoi Intitute of Technology (IIT), Chicago. He i the author of 3 technical paper and four book on electric power ytem planning, operation, and control. Dr. Shahidehpour i an expert in power ytem optimization and control with pecific interet in the modeling of microgrid and utainable energy application. Zuyi Li (M 3,SM 6) received the B.S. degree in electrical engineering from Shanghai Jiaotong Univerity, Shanghai, China, in 995, the M.S. degree in electrical engineering from Tinghua Univerity, Beijing, China, in 998 and Ph.D. degree in electrical engineering from Illinoi Intitute of Technology, Chicago, in. He i currently an Aitant Profeor in the Electrical and Computer Engineering Department and a Reearch Scientit in the Electric Power and Power Electronic Center at Illinoi Intitute of Technology. Hi main reearch interet are market operation of electric power ytem and mart grid.