Testing Market Structures for Electricity Using PowerWeb

Tim MountDepartment of Applied Economics and Management

Cornell University

Ithaca, NY 14853-7801

607-255-4512

TDM2@cornell.edu

Page 2

2

FACULTY PARTICIPANTS Engineers Economists Bob Thomas (Director) Duane Chapman

Jim Thorp Tim Mount Bernie Lesieutre (visiting) Dick Schuler Ray Zimmerman Bill Schulze FINANCIAL SUPPORT - Transmission Reliability Program, US Dept. of Energy - Industrial and Government Members of PSERC

- California Energy Commission

PSERC Research at Cornell University

Page 3

3

Why Use Experiments?

Market structures for electricity auctions are too complicated to derive analytical results.

Experiments are inexpensive compared to experimenting directly on the public.

Paying participants in experiments on the basis of their performance duplicates market behavior effectively.

The effects of specific market characteristics can be isolated and tested.

PowerWeb supports a full AC network, so that the market implications of congestion and ancillary services -- as well as real power -- can be studied.

Our motto: TEST NOW or PAY LATER

Page 4

4

Types of Auction Tested

Uniform Price Auction using the Last Accepted Offer to set the clearing price

Discriminative Auction paying blocks the Actual Offers submitted

Soft-Cap Auction combining a uniform price auction below the cap and a discriminative auction above the cap

Page 5

5

Experiments Using PowerWeb (Spring 2001)

Auctions Tested• Uniform

• Uniform with Price – Responsive Load

• Discriminative

• Soft – Cap

Participants (Representing Suppliers) 1. Cornell University Students (Auctions 1 – 4)

3 Groups of 6 (25 periods)

2. University of Illinois Students (Auctions 1 – 4) 2 Groups of 6 (50 periods)

3. New York Department of Public Services (Auctions 1 and 2) 4 Groups of 6 (30 periods)

Page 6

6

Average Prices for Experiment 1 (uniform)

Page 7

7

Average Prices for Experiment 2 (uniform, price responsive)

Page 8

8

Average Prices for Experiment 3 (discriminative)

Page 9

9

Average Prices for Experiment 4 (soft cap)

Page 10

10

Illustrative Offer Curve for Experiment 1 (uniform)

Page 11

11

Illustrative Offer Curve for Experiment 3 (discriminative)

Page 12

12

Average Prices for High and Low Loads

Page 13

13

Types of Auction Tested Uniform price auction with price inelastic load Soft-cap auction with price inelastic load Soft-cap auction with price responsive load Uniform price auction with price responsive load Objectives Will experts do better than students? Is it practical to run experiments over the internet? Can people exploit a soft-cap auction without experience

in a discriminative auction? Will price responsive load be effective as a way to reduce

prices in a soft-cap auction?

National Experiment (Using PowerWeb)

Page 14

14

National Experiment Analysis of Variance

Main Features 6 Experiments 25 Periods per Experiment 3 Groups of 6 Industry Professionals

Average Price For Last 10 Periods $/MWh Source of Variation Percentage F Statistic

Experiments 73 6.78*Groups 6 1.34Unexplained 21TOTAL 100

• * Statistically Significant

Page 15

15

National ExperimentLegend for Regression Analysis

UN – Uniform Price Auction SC – Soft Cap Auction (Cap at $75/MWh) IN – Inelastic Load PR – Price Responsive Load * * – Initial Costs are High for Marginal Units

UN SC SC**

IN A B E

PR D C F

Page 16

16

National ExperimentRegression Analysis

Average Price $/MWh Variable Coefficient T-Statistic

Mean 67 74.0 Exp.A UN-IN 1 0.5 Exp.B SC-IN -3 -1.5 Exp.C SC-PR -3 -1.6 Exp.D UN-PR -7 -3.7* Exp.E SC-IN** 8 4.3* Exp.F SC-PR** 4 1.9 Group 1 2 1.6 Group 2 -1 1.1 Group 3 -1 0.5 * – Statistically Significant

Page 17

17

National Experiment: Soft-Cap Auction(Average Prices for 3 Groups of Industry Professionals)

Page 18

18

SUNY Binghamton(Course taught my Ed Kokkelenberg)

UN – Uniform Price Auction SC – Soft Cap Auction (Cap at $75/MWh) IN – Inelastic Load PR – Price Responsive Load * * – Initial Costs are High for Marginal Units

UN** SC**

IN A B

PR D C

Page 19

19

SUNY Binghamton: Soft-Cap Auction(Average Prices for 3 Groups of Undergraduates)

Page 20

20

SUNY Binghamton: Uniform Price Auction(Average Prices for 3 Groups of Undergraduates)

Page 21

21

Combined Regression Results Intercept1 82.69 (94.8)*

Experiment2 Difference from the average price in high cost periods3

Changes in price from high cost to low cost periods3

A – S + 11.42 (5.9)* - 7.91 (2.6)* B – S + 4.43 (2.3)* + 6.22 (2.1) B – P - 4.19 (2.2)* - 2.54 (0.8) C – S - 6.29 (3.3)* - 3.52 (1.2) C – P - 6.72 (3.5)* - 5.58 (1.8) D – S + 1.36 (0.7) - 12.15 (4.0)*

1 Average price ($/MWh) in high cost periods with t-ratio in parentheses.

2 A Uniform price auction with inelastic load

B      Soft-cap auction with inelastic load

    C      Soft-cap auction with price responsive load

     D      Uniform price auction with price responsive load

S Students

P Professionals

3 Estimated price change ($/MWh) with t-ratio in parentheses.

* Denotes statistical significance at the 5% level.

Page 22

22

Uniform Price Auction (Pay same price) Infrequent high price spikes are typical Speculating with a FEW units is rational behavior The supply curve looks like a hockey stick Price responsive load mitigates price spikes effectively Discriminative (Soft-Cap) Auction (Pay actual offers) Persistent high prices may occur Speculating with MANY units is rational The supply curve is relatively flat Price responsive load does NOT mitigate high prices

Summary of The Experiments

Page 23

23

Effective countervailing power by loads to mitigate high prices in electricity markets

Effective orchestration of distributed resources for supplying real energy and ancillary services

Active trading of forward contracts in public markets, and better ways to hedge against the

uncertainty of price and load Consistent standards of reporting data to the public Predictability of regulation

Missing Pieces of The Puzzle

Page 24

24

COMING THIS SUMMER TO A PC NEAR YOU

POWERWEB II