Optimal Aggregator Bidding Strategies for Vehicle-To-Grid Energy - - PowerPoint PPT Presentation

1

Optimal Aggregator Bidding Strategies for Vehicle-To-Grid

Energy and the Environment Seminar By Eric Sortomme PhD Candidate, University of Washington October 7, 2010

2

Outline



Introduction



State of the Art in the Field



Research Areas



Preliminary Results

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

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

5

Source: NREL

6

Research Area



The intent of this research is to develop

• ptimal 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.

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

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

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

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

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

• n

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

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

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

15

Unidirectional Regulation Algorithm

MnAPi is the minimum additional power draw

• f the ith EV

PDi is the power draw of the battery of the ith EV CRi is the charge remaining to be supplied to the ith EV MPi is the maximum possible power draw of ith EV MC,i is the maximum charge capacity of the ith EV Efi is the efficiency of the ith EV’s battery charger

Yes

RS >= 0

Yes No

(RS/RD) *MxAPi + POPi < CRi/Efi

PDi = (RS/RU)* MnAPi + POPi

PDi = CRi/Efi

Yes No No

(RS/RU) *MnAPi + POPi < CRi/Efi

PDi = CRi/Efi

PDi = (RS/RD)* MxAPi + POPi

Where: RS is the system regulation signal provided to the aggregator RU is the regulation up capacity of the aggregator RD is the regulation down capacity of the aggregator POPi is the preferred operating point of the ith EV battery MxAPi is the maximum additional power draw

• f the ith

EV SOCi is the state of charge of the ith EV battery

16

Unidirectional Spinning Reserves Algorithm

Yes

RRS >= 0

FPi = (RRS/RR)* RsRPi + PDi

FPi = CRi/Efi

No

(RSS/RR) *RsRPi + PDi < CRi/Efi

Where: RRS is the responsive reserve signal provided to the aggregator RR is the responsive reserve capacity of the aggregator RsRPi is the reduction in power draw available for spinning reserves of the ith EV FPi is the final power draw of the ith EV combining the effects of regulation and responsive reserves

17

Graphical Description

• Reg. Down

Cap.

POP

Maximum Power Draw

• Spin. Res. Signal

Time (min) Battery Power Draw (kW)

• Reg. Up

Cap.

• Spin. Res.

Cap. Actual Power Draw Final Power Draw

18

Graphical Depiction of Variables

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

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

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

22

Price Based Formulation

Max-Actual ( ) Max-Min POP t MP 

5 10 15 20 25 50 100 150 200 250 Time (hours) Day-ahead Price $/MWh Max Actual Min

23

Load Based Formulation

Max-Actual ( ) Max-Min POP t MP 

5 10 15 20 25 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2 2.1 2.2 2.3 x 10

4

Time (hours) Day-ahead Net Load (MW) Max Min Actual

24

MaxReg Based Formulation

( ) Time CR POP t 

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

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

27

Optimal POP Selection (OptComb)

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 SOCI,i is the initial state of charge of the ith EV PRU (t) is the forecasted price of regulation up for time t PRD (t) is the forecasted price of regulation down for time t PRR (t) is the forecasted price of responsive reserves for time t

( ), ( ), ( ), ( )

maximize

i i i i

POP t MxAP t MnAP t RsRP t In

C 

subject to:

( ) ( )

i i

MnAP t POP t i  

 

,

( ( ))

i I i Ci t

E PD t SOC M i   



,

( (1) (1))

i i i I i Ci

MxAP POP Ef SOC M i     ( ) ( ) ( )

i i i

RsRP t POP t MnAP t i    ( ) ( )

i i i

MxAP t POP t MP i    ( )

i

MxAP t i   ( )

i

MnAP t i   ( )

i

RsRP t i   ( )

i

POP t i  

28

Optimal POP Selection

   

( ) ( ) ( ) ( ) ( ) ( ) ( ( ))

RU U RD D RR R t i i t

In P t R t P t R t P t R t Mk E PD t     

 

1

( ) ( )

cars U i i

R t MnAP t



 

1

( ) ( )

cars D i i

R t MxAP t



 

1

( ) ( )

cars R i i

R t RsRP t



  ( ( )) ( ) ( ) ( ) ( )

i i D i i U i R

E PD t MxAP t Ex POP t MnAP t Ex RsRP t Ex    

min min

Pr( )

RS D RS

RS RS dRS Ex RSdRS  



max max

Pr( )

RS U RS

RS RS dRS Ex RSdRS  



max max

Pr( )

RRS R RRS

RRS RRS dRRS Ex RRSdRRS  



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.

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

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 10000 EVs



500 Tesla Roadsters



2000 Th!nk Citys



2500 Mitsubishi i-MiEVs



2000 BMW Mini-Es



3000 Nissan Leafs



Puget Sound commute distance distribution

32

MaxReg Charging Profile Comparison

8 9 10 11 12 13 14 15 16 17 5 10 15 20 25 30 MW (a) MaxReg Algorithm PD POP POP-RUp POP+RDow n 8 9 10 11 12 13 14 15 16 17 5 10 15 20 25 30 Time (hours) (b) MW OptMaxReg Algorithm PD POP POP-RUp POP+RDow n

33

Load Charging Profile Comparison

8 9 10 11 12 13 14 15 16 17 10 20 30 40 MW Load Algorithm (a) 8 9 10 11 12 13 14 15 16 17 10 20 30 40 Time (hours) (b) MW OptLoad Algorithm Max Additional Load POP POP-RUp POP+RDow n Max Additional Load POP POP-RUp POP+RDow n

34

Price Charging Profile Comparison

8 9 10 11 12 13 14 15 16 17 5 10 15 20 25 30 MW (a) Price Algorithm 8 9 10 11 12 13 14 15 16 17 5 10 15 20 25 30 Time (hours) (b) MW OptPrice Algorithm Max Additional Load POP POP-RUp POP+RDow n Max Additional Load POP POP-RUp POP+RDow n

35

Aggregator Profits

$310 $320 $330 $340 $350 $360 $370 $380 Price Load MaxReg OptPrice OptLoad OptMaxReg

Charging Algorithm Profits ($1000)

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

• ptimal charging algorithms significantly outperform

the smart charging analogues



Nevertheless the unconstrained OptMaxReg still yielded the highest profits

37

Daily Peak Load Increase

5 10 15 20 25 P r i c e L

• a

d M a x R e g O p t P r i c e O p t L

• a

d O p t M a x R e g

Charging Algorithm Peak Load Increase (MW) Average Maximum

38

Average Daily Regulation Capacity

0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 8 9 10 11 12 13 14 15 16

Hour Regulation Up Capacity (MW) Price Load MaxReg OptPrice OptLoad OptMaxReg

0.0 5.0 10.0 15.0 20.0 25.0 8 9 10 11 12 13 14 15 16

Hour Regulation Down Capacity (MW) Price Load MaxReg OptPrice OptLoad OptMaxReg

39

Average Energy Price and Minimum SOC (Customer Benefits)

$0.080 $0.082 $0.084 $0.086 $0.088 $0.090 $0.092 $0.094 P r i c e L

• a

d M a x R e g O p t P r i c e O p t L

• a

d O p t M a x R e g

Charging Algorithm Energy Price ($/kWh)

60% 65% 70% 75% 80% 85% 90% 95% 100% P r i c e L

• a

d M a x R e g O p t P r i c e O p t L

• a

d O p t M a x R e g

Charging Algorithm SOC

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.

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

41

Ancillary Services Prices

8 9 10 11 12 13 14 15 16 17 5 10 15 20 25 Time (hours) Ancillary Service Price ($/MW) Regulation Up Regulation Down Responsive Reserves

42

Charging Profiles (OptComb)

8 9 10 11 12 13 14 15 16 17 5 10 15 20 25 MW Time (hours) POP POP-RU POP+RD POP-RU-RR

43

Charging Profiles (OptLoad and OptPrice)

8 9 10 11 12 13 14 15 16 17 5 10 15 20 25 30 35 40 45 MW Time (hours) Max Additional Load POP POP-RU POP+RD POP-RU-RR 8 9 10 11 12 13 14 15 16 17 5 10 15 20 25 30 35 40 45 50 55 MW Time (hours) Max Energy Purchase POP POP-RU POP+RD POP-RU-RR

OptLoad OptPrice

44

Aggregator Profits

$ $1,000 $2,000 $3,000 $4,000 $5,000 $6,000 $7,000 $8,000 OptPrice Opt Load OptComb

Charging Algorithm Profits

45

Regulation Capacities

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 8 9 10 11 12 13 14 15 16 Hour OptPrice OptLoad OptComb 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0 8 9 10 11 12 13 14 15 16 Hour OptPrice OptLoad OptComb

46

Responsive Reserves Capacities

0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0 8 9 10 11 12 13 14 15 16 Hour OptPrice OptLoad OptComb

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

48

Thank you



Questions?