A DISTRIBUTED DEMAND RESPONSE ALGORITHM AND ITS APPLICATIONS TO PHEV CHARGING IN SMART GRID

Zhong Fan

IEEE Trans. on Smart Grid.

Z. Fan. A Distributed Demand Response Algorithm and Its Applications to PHEV Charging in Smart Grid. IEEE Trans. on Smart Gird, vol. 3, num. 3, pp. 1280-1290, 2012.

2

CONTENTS

Demand Response Model

Distributed PHEV Charging

Leveraging Networking Concepts into Smart Grid Load Leveling

3

I - DEMAND RESPONSE (DR) IN SMART GRID

Demand Response (DR): a mechanism for achieving energy efficiency through managing customer consumption of electricity in response to supply conditions.

Ex. Reducing customer demand at critical times (or in response to market price)

Advanced communication will enhance the DR capability (E.g., real-time pricing).

PHEVs require enhanced demand response mechanism.

4

DR MODEL – CONGESTION PRICING

Fully distributed system (only price is known)

A principle of congestion control in IP networks – Proportionally Fair Pricing (PFP)Each user declares a price he is willing to pay

per unit time. The network resource (bandwidth) is shared in

proportion to the prices paid by the users.If each user chooses the price that maximizes

his utility, then the total utility of the network is maximized [1].

[1] F. Kelly, A. Maulloo, and D. Tan, “Rate control for communication networks: Shadow prices, proportional fairness and stability,” J. Oper. Res. Soc., vol. 49, no. 3, pp. 237–252, 1998.

5

DR MODEL AND USER ADAPTION (1)

A discrete time slot system N users demand of user i at slot n user i’s willingness to pay (WTP) parameterPrice of energy in slot n:

Utility function of user i:

The users choose demand to maximize:

6

DR MODEL AND USER ADAPTION (2)

User adaption: user i adapts its demand according to:

The convergence of the adaption:

The error of demand estimate:

: optimal demand

: equilibrium price

8

DR MODEL – NUMERICAL RESULTS (1)

Basic simulation The effect of gamma

9

DR MODEL – NUMERICAL RESULTS (2)

Heterogeneous initial demandsHeterogeneous initial demands and adaption rates

10

DR MODEL – NUMERICAL RESULTS (3)

11

II - DISTRIBUTED PHEV CHARGING

Price function:

User adaption:

Charging dynamics:

Difference: Finish service (Charging done, y=1)

12

DIFFERENTIAL QOS?

Total charging cost for PHEV i:

If we assume the price remains constant (p)

Equilibrium price:

13

DIFFERENTIAL QOS?

Several observationWTPs affect the price of energy.WTPs decide the charging time of individual

PHEVsPHEVs with same total charging demand and

different WTPs will have almost same total charging cost.After some PHEVs finish charging, the price will

go down, which results in slight differences of the charging cost between PHEVs with different WTPs.

14

SIMULATION RESULTS

Basic simulation Differential QoS and total cost of charging Impact of WTPs on system performance Maximum charging rate Different number of PHEVs

15

BASIC SIMULATIONParameter Value

Number of PHEVs 100

Unit of demand 100 kW

Unit of time slot 0.01 h

Initial SOC 15%

Charging efficiency

85%

WTP 0.01+i*0.01

16

DIFFERENTIAL QOS AND TOTAL COST OF CHARGING

Parameter Value

WTP of PHEV50 2, if charging rate <0.2Uniform [0,1], other

WTP of other PHEV Uniform [0,1]

Total charging cost:PHEV1 only 7% less than PHEV50

17

IMPACT OF WTPS ON SYSTEM PERFORMANCE

18

MAXIMUM CHARGING RATEParameter Value

Maximum charging rate 10 kW

WTP Uniform [0,30]

19

MAXIMUM CHARGING RATEParameter Value

Maximum charging rate 10 kW

WTP Uniform [0,30]

20

DIFFERENT NUMBER OF PHEVSParameter Value

Number of PHEVs 20, 60, 100

WTP Uniform [0,2]

21

SOME FUTURE WORK

How should PHEVs adapt their WTPs according to the price policy and their own charging preference?

In-depth analysis of how maximum charging rate impacts the performance.

Game theoretical analysis of the proposed demand response model (the social optimum is a Nash bargaining solution[1])

The impact of PHEVs as energy storage on the SG. The introduction of energy service company (like

charging station) will bring about new problems of optimization, security and social-economic interactions[2].[2] C.Wang and M. de Groot, “Managing end-user preferences in the smart grid,” in Proc. 1st Int. Conf. Energy-Efficient Comput. Network. (ACM e-Energy), 2010.

22

III - INCORPORATING NETWORKING IDEAS AND METHODS INTO THE RESEARCH OF SG

Load leveling as a resource usage optimization problem

Resource allocation ideas from networking to the smart grid.Load admission controlOFDMA allocationCooperative energy trading

S. Gormus, P. Kulkarni, and Z. Fan, “The power of networking: How networking can help power management,” in Proc. 1st IEEE Int. Conf. Smart Grid Commun., 2010.

23

LOAD LEVELING AS A RESOURCE USAGE OPTIMIZATION PROBLEM

Resource allocation:

Optimization goals Environmental impact – load will be shifted to when

the renewable resources have higher general mix. Cheapest resource available – load will be shifted to

the off-peak time when the price is low. When outage?

Hierarchical priority manner Low priority appliances of low priority customer

should be black out first.

24

LOAD ADMISSION CONTROL Like “call admission control” Customers send request before accessing SG to

the Power Management System (PMS) Granted Rejected If the request with high priority

25

OFDMA ALLOCATION OFDMA: deciding which frequencies to

allocate at what times to users Resource allocation in SG: what loads to allocate

at what times to which users to optimize resource utilization and hence improve energy efficiency.

Learn from the OFDMA with the allocation methods

26

COOPERATIVE ENERGY TRADING Future smart grid: micro grids with local

generation plants (solar, wind, etc.) and users while connected to the macro grid.

The idea here is a better utilization of the available power resources by cooperatively using available generation resources.

Similar to the cooperative communication philosophy where the nodes in a wireless network try to increase the throughput and network coverage by sharing available bandwidth and power resources cooperatively.

27

THANKS!