Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Auction designs to increase incentive compatibility and reduce self-scheduling in electricity markets

Conleigh Byers, Brent Eldridge

Abstract

The system operator's scheduling problem in electricity markets, called unit commitment, is a non-convex mixed-integer program. The optimal value function is non-convex, preventing the application of traditional marginal pricing theory to find prices that clear the market and incentivize market participants to follow the dispatch schedule. Units that perceive the opportunity to make a profit may be incentivized to self-commit (submitting an offer with zero fixed operating costs) or self-schedule their production (submitting an offer with zero total cost). We simulate bidder behavior to show that market power can be exercised by self-committing/scheduling. Agents can learn to increase their profits via a reinforcement learning algorithm without explicit knowledge of the costs or strategies of other agents. We investigate different non-convex pricing models over a multi-period commitment window simulating the day-ahead market and show that convex hull pricing can reduce producer incentives to deviate from the central dispatch decision. In a realistic test system with approximately 1000 generators, we find strategic bidding under the restricted convex model can increase total producer profits by 4.4\% and decrease lost opportunity costs by 2/3. While the cost to consumers with convex hull pricing is higher at the competitive solution, the cost to consumers is higher with the restricted convex model after strategic bidding.

Auction designs to increase incentive compatibility and reduce self-scheduling in electricity markets

Abstract

The system operator's scheduling problem in electricity markets, called unit commitment, is a non-convex mixed-integer program. The optimal value function is non-convex, preventing the application of traditional marginal pricing theory to find prices that clear the market and incentivize market participants to follow the dispatch schedule. Units that perceive the opportunity to make a profit may be incentivized to self-commit (submitting an offer with zero fixed operating costs) or self-schedule their production (submitting an offer with zero total cost). We simulate bidder behavior to show that market power can be exercised by self-committing/scheduling. Agents can learn to increase their profits via a reinforcement learning algorithm without explicit knowledge of the costs or strategies of other agents. We investigate different non-convex pricing models over a multi-period commitment window simulating the day-ahead market and show that convex hull pricing can reduce producer incentives to deviate from the central dispatch decision. In a realistic test system with approximately 1000 generators, we find strategic bidding under the restricted convex model can increase total producer profits by 4.4\% and decrease lost opportunity costs by 2/3. While the cost to consumers with convex hull pricing is higher at the competitive solution, the cost to consumers is higher with the restricted convex model after strategic bidding.
Paper Structure (15 sections, 9 equations, 36 figures, 7 tables)

This paper contains 15 sections, 9 equations, 36 figures, 7 tables.

Table of Contents

  1. Introduction
  2. Pricing Models
  3. Methodology
  4. Unit Commitment Model
  5. Nomenclature
  6. Formulation
  7. Offer Strategies
  8. Greedy Algorithm
  9. Exponential Smoothing
  10. Adverse Bidding Test
  11. Illustrative Test Case
  12. Large-Scale Test Case
  13. Data
  14. Results
  15. Conclusion

Figures (36)

  • Figure 1: $|T|=1$, $D_1$ Total actual production cost normalized by production cost at the competitive solution in which all generators bid economically. Higher producer costs indicate the system operator selected a suboptimal solution due to strategic bids.
  • Figure 2: $|T|=1$, $D_1$. Cost to consumers for each pricing model over iterations.
  • Figure 3: $|T|=1$, $D_1$. Percent of iterations that a generator chose each offer strategy.
  • Figure 4: $|T|=1$, $D_1$ Profit duration curve. Generators are sorted by profit achieved in the competitive outcome in which all generators submit economic bids. The mean profit achieved for each strategy in simulation is also shown.
  • Figure 5: $|T|=10$, $D_1$. Percent of iterations that a generator chose each offer strategy.
  • ...and 31 more figures