Optimal Aggregator Bidding Strategies for Vehicle-To-Grid

Transcription

1 Optimal Aggregator Bidding Strategies for Vehicle-To-Grid Energy and the Environment Seminar By Eric Sortomme PhD Candidate, University of Washington October 7,

2 Outline Introduction State of the Art in the Field Research Areas Preliminary Results 2

3 Introduction Electric vehicles (EVs) are poised to receive mass acceptance from the general public Reduced environmental impacts Energy Independence Mass adoption of EVs are not without challenges EVs are more expensive than traditional vehicles Mass, uncoordinated EV charging can cause energy supply issues and distribution overloads One proposed way to address these challenges is through Vehicle-to-Grid (V2G), the provision of energy and ancillary services to the grid from an EV Power flow can be either unidirectional or bidirectional Allows EVs to participate in most energy markets 3

4 Introduction V2G offers benefits to all participants EV owners can generate revenue and receive lower energy prices Utilities can increase system flexibility for increased system control Significant to the operation of V2G are aggregators which bid combined capacities of many EVs into the appropriate markets Base load Peak Energy Spinning Reserves Regulation Non-Spinning Reserves Bidding is not without its challenges EVs can disconnect from the grid whenever the owner has need Owners are not primarily concerned with power market participation and may not do so if there is much inconvenience 4

5 Source: NREL 5

6 Research Area The intent of this research is to develop optimal bidding strategies for aggregators which maximize their profits and the benefits to customers and utilities. If structured correctly, maximum aggregator profits will come from maximum sales of ancillary services which will generate higher revenues for customers and more useful services for utilities. 6

7 Potential Impacts of Research General Welfare Aid in hastening the adoption of EVs Reduced emissions Energy independence Reduced foreign intervention (Mid East Oil Countries) Aggregator Development Provides useful algorithms for different types of controllable loads Utility planning Algorithms can be used by utilities to forecast EV charging impacts and benefits Can determine what load and feeder constraints will need to be imposed 7

8 State of the Art on V2G Early work was feasibility and proof of concept Other recent studies looked at optimizing buying and selling base load and peak energy for single EV The necessity of aggregators to V2G was explored in several recent works. Optimal aggregator bidding of regulation was considered in a recent study 8

9 Unidirectional V2G Most studies considered only bidirectional V2G. Bidirectional V2G has the challenges of: Additional aftermarket hardware for the EVs Interconnection studies and anti-islanding protection at the point of connection Increased cycling wear on the batteries Customer resistance against the idea of letting the utility drain their batteries Manufacturer resistance to utilities degrading their batteries while under warranty. Because of these issues a logical first step is to begin with unidirectional V2G. No aftermarket EV hardware required No interconnection studies nor anti-islanding protection required Battery cycling is not an issue Customer concerns are assuaged 9

10 State of the Art: Unidirectional V2G One study looked at the general concept Involved modulating the charging around the Preferred Operating Point (POP) POP is a market term to describe the operating point scheduled by the aggregator or generator with the system Another study proposed a unidirectional regulation algorithm Dispatches each EV as all on or all off to modulate the group of EVs around the POP, bang-bang charging POP is set to follow renewable energy output levels 10

11 Deficiencies in the State of the Art Ancillary Services Algorithms: Only one regulation algorithm has been proposed Aggregator Bidding Strategies No optimal bidding strategies have been explored Setting the POP has only been lightly touched on 11

12 Research Areas Unidirectional ancillary services algorithms Aggregator profit maximization algorithms for bidding ancillary services and setting the POP for unidirectional V2G Minimization of distribution feeder losses and voltage variations Algorithms for bidirectional V2G 12

13 Unidirectional Regulation Varies each individual EV around its POP to respond to the regulation signal from the aggregator Aggregator capacity is the summation of all individual capacities. Advantages over bang-bang charging are: Reduced the energy losses in the feeders Can operate within a heavily loaded feeder s constraint without overloading the feeder Reduced transients on the distribution system 13

14 Unidirectional Spinning Reserves Similar to regulation but only reduces charging from the POP Can be bid combined with regulation The capacity of regulation up can be broken into regulation capacity and spinning reserves capacity 14

15 Unidirectional Regulation Algorithm RS >= 0 No Yes (RS/RD) *MxAPi + POPi < CRi/Efi (RS/RU) *MnAPi + POPi < CRi/Efi No Yes No Yes PDi = CRi/Efi PDi = (RS/RD)* MxAPi + POPi Where: RS is the system regulation signal provided to the aggregator R U is the regulation up capacity of the aggregator R D is the regulation down capacity of the aggregator POP i is the preferred operating point of the i th EV battery MxAP i is the maximum additional power draw of the i th EV SOC i is the state of charge of the i th EV battery PDi = CRi/Efi PDi = (RS/RU)* MnAPi + POPi MnAP i is the minimum additional power draw of the i th EV PD i is the power draw of the battery of the i th EV CR i is the charge remaining to be supplied to the i th EV MP i is the maximum possible power draw of i th EV M C,i is the maximum charge capacity of the i th EV Ef i is the efficiency of the i th EV s battery 15 charger

16 Unidirectional Spinning Reserves Algorithm RRS >= 0 (RSS/RR) *RsRPi + PDi < CRi/Efi No Yes FPi = CRi/Efi Where: RRS is the responsive reserve signal provided to the aggregator R R is the responsive reserve capacity of the aggregator RsRP i is the reduction in power draw available for spinning reserves of the i th EV FP i is the final power draw of the i th EV combining the effects of regulation and responsive reserves FPi = (RRS/RR)* RsRPi + PDi 16

17 Graphical Description Battery Power Draw (kw) Maximum Power Draw Actual Power Draw POP Final Power Draw Spin. Res. Signal Time (min) Reg. Down Cap. Reg. Up Cap. Spin. Res. Cap. 17

18 Graphical Depiction of Variables 18

19 Aggregator Profit Maximization Any optimization of V2G assets will be done by the aggregator Since the aggregator is a market participant, it is assumed that the aggregator will strive for its own best interests If the markets and regulations are structured properly, this will also lead to maximum benefits for the utilities and customers A fixed percentage of ancillary services revenues and a fixed mark up on energy delivered to the customer accomplishes this objective 19

20 POP Selection The POP will determine how fast the EVs charge and how much regulation capacity can be bid, therefore its selection is very important Two types of POP selection algorithms are considered: Smart Charging Optimal Analogues of the Smart Charging algorithms 20

21 Smart Charging POP Selection Smart charging algorithms considered Price based Load based Considers the net of load and uncontrollable renewables Maximum Regulation Participation (MaxReg) Keeps a near constant POP so as to bid regulation for the entire charging period Does not consider spinning reserves 21

22 Price Based Formulation 250 Max 200 Day-ahead Price $/MWh Actual 50 Min Time (hours) Max-Actual POP() t MP Max-Min 22

23 Load Based Formulation 2.3 x Max Day-ahead Net Load (MW) Actual 1.4 Min Time (hours) POP() t Max-Actual Max-Min MP 23

24 MaxReg Based Formulation POP() t CR Time 24

25 Shortcomings of Smart POP Selection Do not consider price of ancillary services Do not account for aggregator profits Do not view each hour s schedule in terms of the entire charging period 25

26 Optimal POP Selection Maximize the aggregator profits Income comes from a fixed percentage of regulation revenues and a fixed markup on energy Costs are assumed constant since unidirectional V2G has no incremental costs Subject to Maximum and minimum power draw of EV chargers Battery capacities 26

27 Optimal POP Selection (OptComb) maximize POP (), t MxAP (), t MnAP (), t RsRP () t i i i i subject to: MnAPt () POPt () i t i ( ( )), i In C EPD t SOC M i i I i Ci ( MxAP (1) POP (1)) Ef SOC M i i i i I, i Ci RsRP () t POP () t MnAP () t i i i i MxAP () t POP () t MP i i i i MxAPi () t 0 i MnAPi () t 0 i RsRPi () t 0 i POP () t 0 i i Where: In is the income of the aggregator C is aggregator costs Mk is aggregator markup over wholesale energy price α is the percentage of regulation revenue taken by the aggregator SOC I,i is the initial state of charge of the i th EV P RU (t) is the forecasted price of regulation up for time t P RD (t) is the forecasted price of regulation down for time t P RR (t) is the forecasted price of responsive reserves for time t 27

28 Optimal POP Selection In P () t R () t P () t R () t P () t R () t t Mk i cars i 1 cars RU U RD D RR R t E( PD ( t)) RU() t MnAPi() t RD() t MxAPi() t i 1 cars RR() t RsRPi() t i 1 EPD ( ()) t MxAPtEx () POPt () i i D i i MnAP () t Ex RsRP () t Ex i U i R Ex Ex Ex D U R 0 RS min 0 0 RS RRS max RS Pr( RS) drs 0 RS 0 max RS min max RSdRS RS Pr( RS) drs 0 max RSdRS RRS Pr( RRS) drrs RRS RRSdRRS 28

29 Advantages of Optimal POP Selection Makes a plan for each EV for the entire charging period Updates the plan each scheduling period Accounts for ancillary services prices changes over the period If only regulation is bid, the OptComb algorithm is analogous to the MaxReg algorithm and called the OptMaxReg algorithm. 29

30 Adding Load and Price Constraints (OptLoad and OptPrice) A load and price constraints analogous to the smart charging algorithms can be added to the formulation Constrains the aggregator POP to be equal to or less than what would be given under the smart charging algorithms Reduces charging under heavy loading conditions May be mandated by the utility or ISO Reduces customer energy purchases at high prices May be mandated by regulators Gives the aggregator additional freedom however when scheduling individual EVs 30

31 Aggregator Profit Maximization Results Bidding only Regulation Simulated all six POP selection algorithms over an entire year s period Looked only at commuter EVs charging while at work (8AM- 5PM) Test System BPA Wind, Load, Regulation prices for 2007 MidC energy prices for 2007 Hypothetical Group of EVs 500 Tesla Roadsters 2000 Th!nk Citys 2500 Mitsubishi i-mievs 2000 BMW Mini-Es 3000 Nissan Leafs Puget Sound commute distance distribution 31

32 MaxReg Charging Profile Comparison MW MaxReg Algorithm PD POP POP-R Up POP+R Dow n (a) 30 OptMaxReg Algorithm MW PD POP POP-R Up POP+R Dow n Time (hours) (b) 32

33 Load Charging Profile Comparison Load Algorithm Max Additional Load POP POP-R Up MW 20 POP+R Dow n (a) Max Additional Load POP POP-R Up OptLoad Algorithm POP+R Dow n MW Time (hours) (b) 33

34 Price Charging Profile Comparison MW Price Algorithm Max Additional Load POP POP-R Up POP+R Dow n (a) 30 OptPrice Algorithm MW Max Additional Load POP POP-R Up POP+R Dow n Time (hours) (b) 34

35 Aggregator Profits $380 $370 $360 $350 $340 $330 $320 $ OptMaxReg Profits ($1000) Price Load MaxReg OptPrice OptLoad Charging Algorithm

36 Explaining Profits Slide Aggregator markup on energy was high (50 $/MWh) while the percentage of regulation revenues was small (10% of 3.11 $/MWh) For every MWh of energy delivered the aggregator made $50 For every MWh of capacity sold the aggregator made $0.31 With PJM average prices of 35 $/MWh for regulation optimal charging algorithms significantly outperform the smart charging analogues Nevertheless the unconstrained OptMaxReg still yielded the highest profits 36

37 Daily Peak Load Increase Peak Load Increase (MW) Price Load MaxReg OptPrice OptLoad OptMaxReg Charging Algorithm Average Maximum

38 Average Daily Regulation Capacity Regulation Up Capacity (MW) Price Load MaxReg OptPrice OptLoad OptMaxReg Regulation Down Capacity (MW) Price Load MaxReg OptPrice OptLoad OptMaxReg Hour Hour 38

39 Average Energy Price and Minimum SOC (Customer Benefits) $ % Energy Price ($/kwh) $0.092 $0.090 $0.088 $0.086 $0.084 $0.082 $0.080 Price Load MaxReg OptPrice Charging Algorithm OptLoad OptMaxReg SOC 95% 90% 85% 80% 75% 70% 65% 60% Price Load MaxReg OptPrice Charging Algorithm OptLoad OptMaxReg Compared to not using an aggregator, the consumer should benefit. The aggregator markup is over wholesale energy costs, while the consumer standing alone would be paying retail energy costs. Moreover the aggregator is constrained by the retail costs to keep the markup modest. 39

40 Another Case Study Same EV group, same charging times ERCOT system data from the Houston area Simulated over 1 month Prices for ancillary services change each hour Spinning reserves (responsive reserves in ERCOT market) bid Aggregator receives: 10 $/MWh markup on energy 20% of ancillary services revenues Only compared optimal algorithms 40

41 Ancillary Services Prices Ancillary Service Price ($/MW) Regulation Up Regulation Down Responsive Reserves Time (hours) 41

42 Charging Profiles (OptComb) 25 POP POP-R U 20 POP+R D POP-R U -R R 15 MW Time (hours) 42

43 Charging Profiles (OptLoad and OptPrice) MW OptLoad Max Additional Load POP POP-R U POP+R D POP-R U -R R MW Max Energy Purchase POP POP-R U POP+R D POP-R U -R R OptPrice Time (hours) Time (hours) 43

44 Aggregator Profits $8,000 $7,000 $6,000 Profits $5,000 $4,000 $3,000 $2,000 $1,000 $ Opt Price Opt Load Opt Comb Charging Algorithm 44

45 Regulation Capacities OptPrice OptLoad OptComb Hour OptPrice OptLoad OptComb Hour 45

46 Responsive Reserves Capacities OptPrice OptLoad OptComb Hour 46

47 Conclusions Unidirectional V2G can provide significant benefits to utilities, customers, and aggregators Optimal formulations increase the benefits to all participants over heuristic, smart charging methods 47

48 Thank you Questions? 48