The Feasibility of the Ancillary Services for Vehicle-to-Grid Technology

Transcription

1 The Feasibility of the Ancillary Services for Vehicle-to-Grid Technology Siyamak Sarabi 1,, Anouar Bouallaga 1, 3, Arnaud Davigny 1, Benoit Robyns 1, Vincent Courtecuisse 4, Yann Riffonneau 5, Martin Régner 1 LEP-HEI, 13 rue de Toul, 5946, Lille, France, siyamak.sarabi@hei.fr ADEME,, avenue du Grésillé, BP 946, 494, Angers, France 3 SEOLIS, 336 Avenue de Paris, 79 Niort, France 4 GEREDIS Deux-Sèvres, 18 Rue Joule, 79, Niort, France 5 Innovation & Research, SNCF, 4 avenue des Terroirs de France, 75611, Paris, France Abstract This paper explores the feasibility of electric vehicle contribution to the grid ancillary services. Thanks to a probabilistic approach, the technical possibilities of Vehicle-to-grid ancillary services are assessed. Different scenarios with respect to the number of EVs fleet and commuting behavior have been generated to assess the possibility of grid service support in sense of power, energy and appropriate time interval. Each scenario is dedicated to proper ancillary service, where the economic interest for each service has been discussed. Keywords Ancillary services, Distribution Network, Electric Vehicle, Probabilistic algorithm, Vehicle-to-grid (VG). This work is a part of VG project supported half by ADEME (French Environment and Energy Management Agency) and half by SNCF (French National Railway Network), GEREDIS Deux-Sèvres (Distribution System Operator of Deux-Sèvres/France) and SEOLIS company (Energy supplier in Deux-Sèvres /France). I. INTRODUCTION Transportation electrification brings new challenges for the electrical grids. In France, it is assumed to have million electric vehicles by. Without any energy management system, the increasing electricity energy demand could cause negative effects on the grid, especially at distribution grid level. The ability of delivering power from vehicle to the grid, for the first time has been coined Vehicle-to-grid (VG) by AC propulsion Inc. [1]. This ability opened new research horizons to take advantage of EVs increment on the grid. One of these horizons can be assessed as VG grid ancillary services support. In response to the question of Can customers contribute to the ancillary services? P. Sandrin [] says that the customers can act as a generator and also interruptible load, where these contributions should be rewarded for their participation to the grid services. Numerous technical/economical investigations have been conducted to assess the ability and interests of VG ancillary services support [3] [6]. The main services that can be tackled by VG have been considered as regulation, spinning reserve and load leveling, where the regulation services were the most promising services for VG because of high market value and minimum effect on EV storage system [5], [7]. By inspiration from available services for distributed energy storage systems (DESS) [8],[9], the possible services for VG are analyzed in this work. In this paper, thanks to a probabilistic algorithm, different scenarios of EVs fleet in function of numbers and commuting behavior will be generated. After that, with respect to the power and energy, which can be provided by the fleet, the feasibility of VG participation in grid ancillary services will be analyzed. The economic interests for EV owners will be assessed to define the appropriate VG ancillary services in sense of both grid operators and EVs owners. II. VG CONSTRAINTS The participation of EVs in the grid ancillary services requires analytical studies in sense of vehicle technical constraints and possibilities. Although it may well be true that the economic interests of the EV s service integration should be well defined for the vehicles owners, which can be expressed as owners social acceptability. In this study, we mainly focused on technical constraints related to the vehicle s battery where the economic interests also have been introduced. The first component inside the EVs, which will be affected by VG technology, is the battery that accounts more than half of the total vehicle price. There are also constraints related to the chargers, where the most important one is the high efficiency bidirectional charger. For the battery, the technical constraints such as admissible depth of discharge (DOD), cycle life and state of health (SOH) of the battery are the most affected characteristics. Estimating the state of health of the battery needs accurate estimation methods such as discharge tests and ohmic techniques or impulse response analyze [1]. No doubt, the estimation of such characteristics requires deep investigation, which is not included in this study. Instead, we suffice to the previous researches and try to identify the key factors linked with our research aim. Concerning the DOD and its relation with cycle life numerous studies have been done for different applications based on experimental tests [11]. A comparison of DOD vs. Cycle life, for top three contending technologies of EV batteries (Lithium-ion, Lead-acid, Nickel-metal hydride (NiMH)) for deep cycle case (DOD=8%) is depicted in Fig. 1 [1], [13]. The experiments under the full charge testing and various temperatures show the life expectation of 1 years for EV applications and 15 years for HV (Hybrid vehicle) applications for li-ion battery [14] /14/$ IEEE

2 Number of Cycle life NiMH Lithium-Ion Lead-Acid Depth of Discharge % Fig. 1. Cycle life vs. Depth of Discharge comparison Concerning the VG effect on battery life, to respond the question of whether or not participation in VG will also degrade battery life, reference [15] shows half the capacity loss when using batteries for VG compare to rapid driving cycling. For each type of battery, there are some constraints for discharge; NiMH should not be discharged below 8% DOD while Saft Li-ion battery pack can be discharged 1% without any excessive damage [4]. As a reference condition, we have considered DOD=8% where both expected EV battery life and VG service participation will be achievable. The VG ancillary services participation can also be studied as a point of view of the response time of the battery. In [16] the response time for Sodium Sulphur and Lithium-ion batteries has been introduced as in order of few milliseconds. More accurately, the response time of milliseconds is claimed for Lithium-ion battery in [17].This range is also true in Saft Lithium-ion technology for both charging and discharging periods [18]. This relatively high response action can be considered as one of the advantages of battery energy storage systems. For the different grid services, the range of seconds to minute s response time is a key factor and this feature of battery energy storage systems makes them valuable in sense of grid service participation. The other constraints that affect the feasibility of grid service participation are mainly related to the stochastic behavior of the EVs fleet. The information such as commuting time of the vehicles, available state of charge and unforeseen cases that affects the availability of the vehicles, is the main stochastic data. For this study, we have designed an algorithm based on probabilistic approach that will generates the different scenarios based on available state of charge, arrival and departure time and daily driving distance. Thanks to the provided information, the available VG energy for each vehicle will be calculated. These generated scenarios provide a benchmark, which facilitate the grid support assessment of the EVs fleet. A. Probabilistic algorithm In this research, we have focused on French electric vehicle market and the EV production between the years 1 to 13. The proposed algorithm with inspiration from [19] generates a fleet of EVs based on information (e.g., battery capacity, daily driving distance and commuting times). This algorithm consists of 5 steps explained precisely thereafter. 1) EV type and capacity: in the first step, the algorithm chooses the type of each EV in the fleet based on the market contribution portion available in []. The selection is random but with the probability of market contribution for each type. By choosing the EV type the other characteristics such as, battery capacity, battery type and NEDC (New European Driving Cycle) autonomy will be provided. ) Driving distance (D d): the information related to the daily driving distance has been considered based on 3 km average daily home/work distance in France [1]. The algorithm based on uniform distribution dedicates randomly driving distance, since the battery was fully charged, D d [1 km, 5km], to the every single vehicles. 3) Commuting times: based on the traffic habits of Île-de- France region in France, during the working days (Monday to Friday), sharp increment of vehicles commuting around 8: AM and dramatically increment during 6: PM happen. This is the same for Friday, where a bigger peak happens at 6: PM. In this study, we have considered a fleet of vehicles which are commuting regularly between homes to work during the working days. To generate the set of arrival and departure time for the vehicles, two normal distributions with average values of μ=18: for arrival time (AT) and μ=8: for departure time (DT) have been generated. For standard deviation in each case the two values of =1 and =3 hours have been considered to cover both case of commuting behavior during the week. 4) Arrival SOC (State of Charge): the assumption in this study is that the all EVs are used by commuting purpose and every EV will be charged and discharged once in a day, where the VG application is also included. The available SOC at the moment of arrival (SOCarrival), by assuming linearly drop with the distance of travel, can be calculated as in (1): Dd SOCarrival 1 1% A (1) Where A is the NEDC autonomy of the vehicle in km. 5) VG capacity: at this step, for each vehicle arriving home, the participating capacity in VG grid services can be calculted, where the EVs first target should be respected. This means to provide fully charged battery at departure time. Dd EV G( DOD EEV ) C () Equation (), calculates the available VG energy, E VG in the period of home parking. EEV indicates the battery capacity in kwh, EV denotes the vehicle deriving efficiency in km/kwh which in this study, because of lack of information, has been considered as NEDC autonomy divided by battery capacity. C is also the charger efficiency which is considered as.97 according to [19]. To respect the EVs owners priority the constraint (3) should be considered. D E T P (3) EV d V G Plug _ in outlet EV

3 Where P outlet, denotes the charging/discharging power, which has been considered as 3kW for the case of normal charging rate, and T plug-in the plug-in time at home. This constraint ensures the fully charged battery at the departure time even after VG application. The output of whole five steps of the algorithm for a case of 5 EVs fleet can be found in Fig.. From the results of this part we will define different scenarios to provide the benchmark for the grid services support. B. Scenario generation In this part, 8 scenarios for different numbers of EVs and different commuting behaviors thanks to the previous algorithm will be generated. The results of generated scenarios are presented in Table I. For each scenario minimum available arriving number (MA) of EVs per minute with its availability factor in the interval of [μ-, μ+] will be calculated. This interval has been chosen as a potential grid support period and will be called as potential interval. In this interval, the minimum available power deduced from MA is called as Minimum VG Power (MVP) which is considered as an indicator for each scenario. This minimum power is equal to 447 kw for 5 EVs fleet demonstrated in Fig. 3. EV ( ) (,, ) nb x f x n (4) ( x ) 1 f( x,, ) e (5) kx ( ) 1 EVnb ( x) EVnb ( ), x[, ] (6) mx ( ) 1 EVnb ( x) EVnb ( ), x[, ] kx ( ) AF 1 xmx ( ) kx ( ) (7) Where in (4), EVnb ( x ) gives the number of EVs in time x, as a function of probability distribution function f( x,, ) and total number of fleet, n. kx ( ) and mx ( ) will work as counters of minutes, when the number of EVs per minute are respectively bigger and smaller than minimum numbers per minute in the potential interval. Finally in (7), the availability factor will be calculated in the potential interval. An example for case of 5 EVs has been presented in Fig. 3. This benchmark can be applied to VG ancillary service analyzes, where each scenario presented in Table I, will be examined for appropriate ancillary services. The focus of this study is on the services that can be tackled on Medium Voltage (MV) and Low Voltage (LV) distribution network level. III. ANCILLARY SERVICES (AS) For the sake of maintaining grid reliability and balance of production and consumption, the existence of ancillary services are necessary. These services support the reliable transmission of power from producers to the customers [].With respect to the commute behavior of transportation fleet, the majority of EVs are used for daily home/work round trip. Average 3 km daily trip make them available and parked near to 95% of the time. Also based on available arrival SOC (Fig. ) for a case of normal charging (3 kw) a single EV needs 1 and half hour average charging time. This makes it fully charged and uselessly plugged, in most of the plugged-in time. For this reason, the EVs fleets have high potential for grid service support in these periods. Number of arriving EVs EV type distribution Battery capacity (kwh) Driving distance uniform distribution Round trip driving distance (km) Arrival SOC distribution Time (hour) VG energy distribution Arrival SOC %.6.4. Arrival time histogram Departure time histogram VG capacity (kwh) Fig.. The five steps outputs of probabilistic algorithm for 5 EVs Normal distribution of arrival time ( =1 hour, = 18:) 5 Minimum arriving EV ( nb - ) =1 AF = 95% Arrival time Potential Interval [-, +] Time (hour) Cumulative number of arriving EVs Cumulative distribution function CDF Potential Interval [-, +] Mean power =7695kW Minimum Power =447kW Time (hour) Fig. 3. Example of 5 EVs fleet, calculation of availability factor, minimum and average potential power TABLE I. Scenario GENERATED SCENARIOS BY PROBABILISTIC ALGORITHM VG Power and Energy Potential * EV h EV/ AF min % kw kw MWh AT DT PI MA MVP AVP VE S S S S S S S S * EV: Electric Vehicles numbers, AT: Arrival Time, DT: Departure Time, PI: Potential Interval, MA: Minimum Arriving, MVP: Minimum VG Power, AVP: Average VG Power, VE: VG Energy, AF: Availability Factor

4 In this section analyses about the competitive VG ancillary services, by inspiration from DESS ancillary services for distribution MV and LV grid provided in [9], will be conducted. Therefore, based on the previous obtained scenarios, the competitive services, considering the DESS services requirement (Table II), will be assessed. A. Regulation Regulation consists of the processes which lead to balance the production and consumption in the whole electricity grid to achieve fine-tune of frequency of the grid. In France the Réseau de Transport d Electricité (RTE), which is part of the European Network of Transmission System Operators for Electricity (ENTSO-E), is acting as a transmission system operator (TSO), responsible for regulation services. This service consists of primary, secondary and tertiary control. 1) Primary control (AS1): Primary control is acting automatically within 15 to 3 seconds to stabilize the frequency at a reference value of 5 Hz. This regulation is common (i.e. similar to PJM frequency response) for all the ENTSO-E members with a 3 MW capacity [6]. All generating units with a capacity higher than 4 MW have to contribute to primary control. This represent 6 MW for the French system. In France, the primary reserve receives a fixed capacity payment, i.e. 8 /MW within 3 minutes [6]. The ability of an aggregated EV fleet providing primary control has been assessed in [6] where the results showed contribution to funding energy cost and battery investment cost reduction. For this service, comparying with the generated scenarios, the requirements is met by scenarios S7 and S8 with minimum VG power of.7 and.4 MW (atleast 5 EVs fleet), respectively. Controlling these amount of EVs needs aggregation strategies at TSO level. ) Secondary control (AS): The secondary reserve will be activated automatically after primary within 1 to seconds. All producers in France with more than 1 MW capacities are required to allocate a portion to the secondary reserve to provide total capacity between 5 to 1 MW for secondary reserve. Secondary reserve makes possible the restoring for primary control, where tertiary makes it for secondary. Secondary reserve is paid for capacity, the same as primary, and for energy delivered, (i.e. 9.3 /MWh [6]). In this service, S7 and S8 are more competitive than others where the same control strategies are essential. For the EV owners secondary reserve is more interesting since they will be paid for both energy and capacity. 3) Tertiary control (Balancing mechanism)(as3): The French producers and customers, subject to the availability of 1 MW, are encouraged to participate in adjustment mechanism (balancing mechanism). This bid will be additional to the contracted reserves. Balancing mechanism is a part of tertiary control of the regulation services which is called as 3 minutes complementary tertiary reserve [3]. In this service, Controlling the EV fleet is much more difficult as the bidding capacity is quite larger. Although, the interests for capacity payment increase the EV owners participation. B. Peak power shaving(as4) For the DESS application, peak shaving and valley filling is when a storage unit is used to shift the load from peak to offpeak hours. Thanks to a supervision system, by controlling the charging periods of EVs fleet and shifting as often as possible the charging time to the off-peak hours, the number of subscribed power violations (SPV) will be reduced. Therefore, the interests are: The investment for grid infrastructure reinforcement will be postponed (e.g. investment reduction up to 35% for distribution networks [4]). The energy transmission cost minimization; In France, distribution system operator (DSO) pays to TSO for its energy transmission cost which highly depends on SPV and extracted energy. As a point of view of an energy supplier, for the sake of energy purchase price optimization, the customers should be encouraged to participate in peak shaving service by consuming more during low-cost periods and less during high-cost hours. This could be an opportunity for VG enabled EVs to contribute to the grid ancillary services. The customers (e.g. EV owner) by participating in this service can minimize the energetic part of their invoice (i.e. maintain at the level it was before purchasing an EV) as well as gain from VG services. This participation in case of predictable peaks and short period can be profitable (e.g. up to 1 /kw [9]). By comparing the scenarios with requirements, for this service, Scenario S6 with 54 kw power for MV level, and S3 and S4 for LV level can be considered. A supervision system, in DSO) level, can properly manage a fleet of 5 to 1 EVs to participate in peak shaving service. C. Reactive power compensation (AS5) As a part of power quality services, thanks to the power electronics converters, the reactive power correction can be tackled by VG by controlling the angle between voltage and current without any major effect on the charging process. This can be achieved in both charging and discharging periods. For this service, S3 and S4 (5 EVs fleet) can be sufficient. D. Renewable energy support (AS6): The variable behavior of renewable energy sources (e.g. wind and solar) reduces their contribution to the grid service as there is no high reliability to their availability. Thanks to the storage systems, their contribution to the ancillary services, such as primary control will be increased. In this case the Distributed Generation (DG) producers are able to provide reliable control power to the grid. VG fleet can act as a DESS to support the storage need of the DG centers by coordination of its charging time with the production of DG center. The only thing that may reduce the interests for EV owners is the stochastic behavior of such sources, when the proposed charging periods may not be necessarily at the availability periods of the EVs.

5 E. Base load (AS7) Base load power is provided continuously by large generation unit which has low cost per kwh. VG has been analyzed for this service [3], [4] showing that there is no competitive price as the EVs have limited storage capacity, high energy cost per kwh and less charging access at transmission level. Ancillary Service TABLE II. Minimum Required power DESS ANCILLARY SERVICE REQUIREMENTS DESS services requirement Benefitted stakeholder Appropriate VG scenario Required Interval AS1 1 MW 3 sec-15 min TSO S7-S8 AS -3 MW 3 sec ALAR a TSO S7-S8 AS3 1 MW 3 min 6 h TSO Several S8 AS4 5 kw (MV) TSO-DSO- -1h 1 kw (LV) Costumer S3-S4 AS5 1 kvar ALAR DSO S3-S4 AS6 1kW-MW min-1h3 Distributed generators S3 to S8 ALAR: As Long As Required IV. CONCLUSION In this study, the possibility of VG participation in grid ancillary services has been assessed. The different technical constraints of the electric vehicles have been analyzed to identify the limitation of VG ancillary services. These limitations consist of EVs availability and admissible depth of discharge of 8%, where both expected EV battery life and VG service participation will be achievable. In addition, thanks to a probabilistic algorithm, the availability limitation of EV fleet has been discussed and different scenarios, according to the number of EVs and commuting behaviors of the fleet, have been generated. The ancillary services competitive for DESS systems have been introduced. By considering the EVs fleet as a distributed energy storage unit, each scenario has been dedicated to the appropriate service. As a conclusion, the minimum fleet of 5 EVs is necessary for service participation (Scenarios S1 and S are not sufficient). The peak power shaving service can be competitive for the EVs fleet, where the interests cover the majority of stakeholders (e.g. TSO, DSO, energy provider and customers). In addition, possibility to participate in regulation services (especially secondary and tertiary) increase the economic benefits for both providers and customers as there is extra income related to the bidding capacity. REFERENCES [1] A. Emadi, Handbook of Automotive Power Electronics and Motor Drives. 5. [] P. Sandrin, System-services: several questions, in an integrated utility perspective [ancillary services], in Pricing of Ancillary Services: an International Perspective (Digest No: 1996/164), IEE Colloquium on, 1996, pp [3] -to-grid power implementation: From stabilizing the grid to supporting large-scale renewable energy, J. Power Sources, vol. 144, no. 1, pp. 8 94, Jun. 5. [4] -to-grid power fundamentals: Calculating capacity and net revenue, J. Power Sources, vol. 144, no. 1, pp , Jun. 5. [5] Z. Luo, Z. Hu, Y. Song, Z. Xu, H. Liu, L. Jia, and H. Lu, Economic analyses of plug-in electric vehicle battery providing ancillary services, in Electric Vehicle Conference (IEVC), 1 IEEE International, 1, pp [6] M. Petit and Y. Perez, Vehicle-to-grid in France: What revenues for participation in frequency control?, in European Energy Market (EEM), 13 1th International Conference on the, 13, pp [7] -drive vehicles for grid support, J. Power Sources, vol. 168, no., pp , Jun. 7. [8] A. Oudalov, D. Chartouni, C. Ohler, and G. Linhofer, Value analysis of battery energy storage applications in power systems, in Power Systems Conference and Exposition, 6. PSCE 6. 6 IEEE PES, 6, pp [9] G. Delille, B. Francois, G. Malarange, and J.-L. Fraisse, Energy storage systems in distribution grids: new assets to upgrade distribution networks abilities, in Proc. th International Conference and Exhibition on Electricity Distribution (CIRED9), 9. [1] A. Banaei and B. Fahimi, Real time condition monitoring in Li-Ion batteries via battery impulse response, in Vehicle Power and Propulsion Conference (VPPC), 1 IEEE, 1, pp [11] T. Guena and P. Leblanc, How depth of discharge affects the cycle life of lithium-metal-polymer batteries, in Telecommunications Energy Conference, 6. INEC 6. 8th Annual International, 6, pp [1] M. Duvall and M. Allan, Batteries for Electric Drive Vehicles Status 5., Electric Power Research Institute, Tech. Rep., 5. [13] H. Bindner, T. Cronin, P. Lundsager, Risø National Lab, and Roskilde (DK). Wind Energy Department, Lifetime modelling of lead acid batteries. 5. [14] G. Sarre, P. Blanchard, and M. Broussely, Aging of lithium-ion batteries, J. Power Sources, vol. 17, no. 1, pp , Mar. 4. [15] S. B. Peterson, J. Apt, and J. F. Whitacre, Lithium-ion battery cell degradation resulting from realistic vehicle and vehicle-to-grid utilization, J. Power Sources, vol. 195, no. 8, pp , Apr. 1. [16] M. C. Kintner-Meyer, P. J. Balducci, C. Jin, T. B. Nguyen, M. A. Elizondo, V. Viswanathan, X. Guo, and F. Tuffner, Energy Storage for Power Systems Applications: A Regional Assessment for the Northwest Power Pool (NWPP)., vol Pacific Northwest National Laboratory, 1. [17] Li-ion Battery - Centre for Low Carbon Futures, Lithim-ion battery. [Online]. Available: pdf. [18] Saft Lithium-ion battery. [Online]. Available: [19] H. Turker, S. Bacha, D. Chatroux, and A. Hably, Modelling of system components for Vehicle-to-Grid (VG) and Vehicle-to-Home (VH) applications with Plug-in Hybrid Electric Vehicles (PHEVs), IEEE PES Innov. Smart Grid Technol. ISGT, pp. 1 8, 1. [] S. Sarabi, L. Kefsi, A. Merdassi, and B. Robyns, Supervision of Plug-in Electric Vehicles Connected to the Electric Distribution Grids, Int. J. Electr. Energy, vol. 1, no. 4, pp , 13. [1] The commuting amplified by suburbanization, DADS studies, National Institute of Statistics and Economic Studies in France, 4. [Online]. Available: [] M. Yilmaz and P. T. Krein, Review of the Impact of Vehicle-to-Grid Technologies on Distribution Systems and Utility Interfaces, IEEE Trans. Power Electron., vol. 8, no. 1, pp , Dec. 13. [3] Y. Rebours and D. Kirschen, A Survey of Definitions and Specifications of Reserve Services, Univ. Manchester, Tech. Rep., 5. [4] L. Pieltain Fernandez, T. Gomez San Roman, R. Cossent, C. Mateo Domingo, and P. Frias, Assessment of the Impact of Plug-in Electric Vehicles on Distribution Networks, IEEE Trans. Power Syst., vol. 6, no. 1, pp. 6 13, Feb. 11.