Bidding Rules in Deregulated DayAhead Electricity Markets

1
Bidding Rules in Deregulated Day-Ahead
Electricity Markets

• Deddy Koesrindartoto

2
Introduction

• Electricity deregulation motivation
• Excess supply capacity
• Vertically integrated
• Market power

• Regulator strategy
• Separate vertically integrated market
• Introduce competitive bidding
• Provide missing markets

April 2003 US Federal Energy Regulatory
Commission proposed Wholesale Power Market
Platform (WPMP) dubbed as Standard Market Design
(SMD) for common adoption by US wholesale power
market.
3
Wholesale Power Market Components

• SMD meets basic WPMP structural requirements
• Independent System Operator (ISO)
• Day-ahead and real-time electricity markets
• Congestion managed via LMP
• Financial Transmission Rights
• Planned reserve and capacity markets as well as
  enhanced demand response

4
Simplified SMD Activity Flow
5
(000-2300)
(1600)
(000-2300)
(000-2300)
(000-1200)
(000-1200)
(1600-1800)
(000-2300)
Real-Time
Daily
Monthly
(000-1200)
FTR
D1 Day-Ahead
(1600-1800)
D1 Supply Re-Offers
6
Focus of this Study

• This study is motivated by findings of Klemperer
  and Meyer (1989) and subsequent related works on
  the single-sided seller auction.
• This previous work found it was beneficial to
  sellers to submit their supply offers in
  functional P(Q) form rather than in point (P,Q)
  form. This study shows that this finding is not
  robust across different market conditions.
• This study also investigates the possible
  benefits to both buyers and sellers of submitting
  their bids and offers in functional P(Q) form
  rather than point (P,Q) form in a double auction.
• To make this study more complete, this study try
  to replicate Rassenti, Smith and Wilson (2000)
  human-subject experiment to analyze the benefit
  of switching from a single-side auction to a
  double-sided auction using point (P,Q) bid rule.
  This study shows that this finding is robust
  across different market conditions.
• As will be shown below, in the double auction it
  is generally not beneficial either to sellers or
  to buyers to submit their bids and offers in
  functional P(Q) form rather than point (P,Q) form.

7
Focus of this Study
continued
? Rassenti et al. (2000)
? Koesrindartoto (2005)
? Klemperer and Meyer (1989)
8
Key Findings

• Point (P,Q) bid persistently results better
  market outcomes low market clearing-prices, high
  efficiency, and persistent market mitigation.
• On the other hand, function P(Q) bid persistently
  cause coordination failures that can decreases
  the market efficiency and reduce the benefits of
  a double-auction market.
• Shows the robustness Rassenti, Smith and Wilson
  (2000) findings under different market
  treatments.
• Difficulties to replicate Klemperer and Meyer
  (1989) results even under similar market
  environments.

9
The Model

• There are S sellers and B buyers actively engaged
  in a sealed-bid multi-unit double auction.
• Each seller s has a maximum supply capacity Qs
  and a constant marginal cost MCs
• Each buyer b has a maximum demand capacity Qb
  and a constant marginal revenue MRb
• Each seller/buyer learns via Roth-Erev
  reinforcement learning
• Auctioneer calls and clears sellers/buyer
  offers/bids under different bidding rules
• (P,Q) point bidding rule
• Each buyer submits (pb,qb ) with 0lt pb ?
  MRb and qb Qb
• Each seller submit (ps,qs) with ps MCs
  and qs Qs
• P(Q) functional bidding rule
• Each buyer submits pb(q) ab bb.q , q
  1,2, .. Qb , ab gt 0, bb 0
• Each seller submits ps(q) as bs.q , q
  1,2, .. Qs , as gt 0, bs 0

10
The Model (continued)

• Key market outcomes ( market efficiency, market
  power, cleared quantities, sellers and buyers
  profits) are compared under two different market
  clearing-rules uniform-price vs.
  discriminatory-price.
• For completeness, as in Klemperer and Meyer
  (1989), experiments are also conducted for a
  single-sided auction with a given aggregate
  demand function for buyers and with sellers
  submitting both functional offers P(Q) and point
  offers (P,Q)

11
Experimental Design Parameters

• Bidding rule
• point (P,Q) bidding rule vs. function P(Q)
  bidding rule.

2. Trading rule active vs. passive demand
bidding (double-sided
auction vs. single-sided auction)
3. Market structures combination of relative
capacity and relative concentration

Total Demand Capacity Relative Capacity
--------------------------------
- (RCAP)
Total Supply Capacity

Total Number of Sellers S Relative
Concentration ------------------------
-------------- (RCON)
Total Number of Buyers B
12
Fig 2. Geometric representation of the
experimental design used for each
market-clearing rule
13
Experimental Design Parameter Values
a. Market Structures
S of sellers B of
buyers Qs individual seller capacity Qb
individual buyer capacity
14
Fig 3. True demand and supply curves with
different (RCONRCAP) values
15
Experimental Design (continued)

• Benchmark
• Sellers and Buyers submit their
  actual reservation values.
• Experiment 1A (single-sided auction)
• Sellers submit point offers (P,Q)
• Buyers submit passive demand bids (fixed demand
  function)
• Experiment 1B (double-sided auction)
• Sellers and buyers submit point offers and bids
  (P,Q)
• Experiment 2A (single-sided auction)
• Sellers submit functional supply offers P(Q)
• Buyers submit passive demand bids (fixed demand
  function)
• Exeperiment 2B (double-sided auction)
• Sellers and buyers submit functional offers and
  bids P(Q)

16
Market Performances

• Market-clearing price
• Market efficiency
• how good is the market design in generating
  total
• surplus.
• Market power mitigation
• how good is the market design mitigating
  traders market
• power.

17
Fig 4. Market-clearing prices with different
(RCONRCAP)
SA vs DA under (P,Q) bidding rule
SA vs DA under (P,Q) bidding rule

• Note
• Benchmark Competitive auction (bids/offers at
  reservation values)
• E1A Experiment 1A, single-sided auction with
  (P,Q) point bidding by sellers
• E1B Experiment 1B, double auction with (P,Q)
  point bidding by buyers and sellers
• E2A Experiment 2A, single-sided auction with
  P(Q) functional bidding by sellers
• E2B Experiment 2B, double auction with P(Q)
  functional bidding by buyers and sellers

18
Fig 5. Market efficiencies with different
(RCONRCAP) values

• Note
• Benchmark Competitive auction (bids/offers at
  reservation values)
• E1A Experiment 1A, single-sided auction with
  (P,Q) point bidding by sellers
• E1B Experiment 1B, double auction with (P,Q)
  point bidding by buyers and sellers
• E2A Experiment 2A, single-sided auction with
  P(Q) functional bidding by sellers
• E2B Experiment 2B, double auction with P(Q)
  functional bidding by buyers and sellers

19
Fig 6. Sellers profit with different
(RCONRCAP) values

• Note
• Benchmark Competitive auction (bids/offers at
  reservation values)
• E1A Experiment 1A, single-sided auction with
  (P,Q) point bidding by sellers
• E1B Experiment 1B, double auction with (P,Q)
  point bidding by buyers and sellers
• E2A Experiment 2A, single-sided auction with
  P(Q) functional bidding by sellers
• E2B Experiment 2B, double auction with P(Q)
  functional bidding by buyers and sellers

20
Fig 7. Buyers profit with different (RCONRCAP)
values

• Note
• Benchmark Competitive auction (bids/offers at
  reservation values)
• E1A Experiment 1A, single-sided auction with
  (P,Q) point bidding by sellers
• E1B Experiment 1B, double auction with (P,Q)
  point bidding by buyers and sellers
• E2A Experiment 2A, single-sided auction with
  P(Q) functional bidding by sellers
• E2B Experiment 2B, double auction with P(Q)
  functional bidding by buyers and sellers

21
Fig 8. Market-clearing quantities with different
(RCONRCAP) values

• Note
• Benchmark Competitive auction (bids/offers at
  reservation values)
• E1A Experiment 1A, single-sided auction with
  (P,Q) point bidding by sellers
• E1B Experiment 1B, double auction with (P,Q)
  point bidding by buyers and sellers
• E2A Experiment 2A, single-sided auction with
  P(Q) functional bidding by sellers
• E2B Experiment 2B, double auction with P(Q)
  functional bidding by buyers and sellers

22
Fig 9. Aggregate sellers market power with
different (RCONRCAP) values

• Note
• Benchmark Competitive auction (bids/offers at
  reservation values)
• E1A Experiment 1A, single-sided auction with
  (P,Q) point bidding by sellers
• E1B Experiment 1B, double auction with (P,Q)
  point bidding by buyers and sellers
• E2A Experiment 2A, single-sided auction with
  P(Q) functional bidding by sellers
• E2B Experiment 2B, double auction with P(Q)
  functional bidding by buyers and sellers

23
Fig 10. Aggregate buyers market power with
different (RCONRCAP) values

• Note
• Benchmark Competitive auction (bids/offers at
  reservation values)
• E1A Experiment 1A, single-sided auction with
  (P,Q) point bidding by sellers
• E1B Experiment 1B, double auction with (P,Q)
  point bidding by buyers and sellers
• E2A Experiment 2A, single-sided auction with
  P(Q) functional bidding by sellers
• E2B Experiment 2B, double auction with P(Q)
  functional bidding by buyers and sellers

24
Conclusions

• Point (P,Q) bid persistently results better
  market outcomes low market clearing-prices, high
  efficiency, and persistent market mitigation.
• On the other hand, function P(Q) bid persistently
  cause coordination failures that can decreases
  the market efficiency and reduce the benefits of
  a double-auction market.
• Shows the robustness Rassenti, Smith and Wilson
  (2000) findings under different market
  treatments.
• Shows the Klemperer and Meyer (1989) result is
  not robust even under similar market
  environments.

25

• The end