ECF 5360 Final Presentation

2026-05-18

Overview

Short Term Market Design of wholesale markets and Physical reality of grids: Divergence and Impacts
One chocolate to be earned
Two papers to cover:
- How new (and old) power system operating constraints map to Australia’s wholesale electricity market model by Dr. Leslie and Dr. Billimoria
- Rethinking wholesale market design for New Zealand’s clean energy transition by Dr. McRae

Why (Motivation)

System security directions in South Australia have increased in cost and frequency.

In New Zealand, August 2021, the system operator declared a grid emergency, leading to rolling blackouts that affected 34,000 customers.

Power system physics and the market model

For Australia
- Zonal Pricing
- Static Loss Factors
- Fixed Retail Rates
For New Zealand
- Single-part bidding
- Single-settlement Market Structure
However, Australia also has a single settlement system!

Australia: Case 1

Generator A and B have an incentive to bid at floor price ($-1,000)

Australia: Case 1

Real World Problems
- Katzen and Leslie (2024) measure “mispriced generator revenues” by comparing them to hypothetical applied LMPs. Mispriced generator revenue increased from 0 (2013) to $358m in 2019 (2.2% of all wholesale revenues)
  - Intermittent Renewables have correlated, low-cost output located away from load centers
- Snowy has been documented in congested situations to update both it’s supply offers and it’s unit-level ‘ramp down’ rates to the minimum allowed 3 MW/min to maximise output.
  - Storages are likely to have similar Market Power, which will make the problem worse

Australia: Case 2

Storages operate on the “buy low-sell high principle”
In Australia, two pricing rule elements impede this strategy:
- Storages face the RRP while charging, not the $0/MWh which is the “real” price of this renewable output which would otherwise be curtailed
- Loss factors for each unit in the NEM are set anually. Actual loss factors are dynamic and volatile. Storage reduces actual loss factors for all units but this benefit goes unpriced.

Australia: Case 3

The NEM outlines minimum unit configurations for all mainland states to ensure system security
NEMDE does not solve a mixed-integer program to identify the least-cost combination of relevant plants to maintain a secure system
Units can be “directed” to run by the NEM. If directed, they receive the 90th percentile of spot prices over the trailing 12-month window and any required additional compensation

Australia: Case 3

If firms might reasonably believe that:
- System Security conditions will not be met
- They are pivotal in the conditions being met
It is in their private economic interest to appear uneconomic in NEMDE (via their bids) and to instead be directed on and receive the directed price
South Australia operated under directions in over 60% of market intervals in Q4 2023 and Q1 2024
AEMO observed that “directed” generator output in South Australia halved from Q2 2023 (compensation payments were $337/MWh) to Q2 2024 (compensation payments were $179/MWh)
Intermittent sources do not provide system security, which means this problem will become increasingly important

Australia: Case 4

Operational demand is very low at the peak of rooftop solar production
Currently, this can be used to serve other regions but eventually this ability to evacuate the power would be limited.
The system needs to curb such generation and encourage incremental demand at these times

Australia: Case 4

However,
- Energy consumers are subject to primarily fixed retail rates. This leads to misalignment between consumption decisions and system needs.
- Rooftop generation is paid via feed-in-tarrif schemes that similarly do not adjust with wholesale prices.
Since these higher weighted wholesale prices are atleast somewhat passed on to fixed retail tarrifs, there is implicit cross-subsidisation. A 2023 study analysing Victoria, ceteris parabus, found a 10% difference in the average wholesale procurement cost for households in the highest rental neighborhoods and the highest owner-occupier neighborhoods
The solution may lie in real time picing, maybe with insurance against extremely high price shocks, or even just time-of-use pricing
Distributed Energy Resources suffer from double coincidence of wants, so the second point might not have the highest impact.
In April 2026, AEMO is proposing to temporarily pause and reset the delivery path for the Integrating Price Responsive Resources into the NEM (IPRR) rule implementation.

New Zealand: Unit-Commitment Issues

Total Demand is 100 MW
Value of Lost Load, which represents the economic cost of involuntary load shedding is $10,000/MW

Table 1: Characteristics of generation plants
Plant	Capacity (MW)	Marginal Cost	Startup Cost ($)	Lead Time (Periods)
Wind (High)	60 (p = 0.8)	0	0	0
Wind (Low)	20 (p = 0.2)	0	0	0
Baseload	70	30	1000	1
Peaker	70	100	0	0

New Zealand: Socially optimum combination

So, what would a social planner minimizing the expected costs want as the generation combination?
In high-wind scenario:
- Wind for 60 MW + Baseload for 40 MW
- The total cost is (0 × 60) + 1000 + (30 × 40) = $2,200
In low-wind scenario:
- Wind for 20 MW + Baseload for 70 MW + Peaker for 10 MW
- The total cost is (0 × 20) + 1000 + (30 × 70) + (100 × 10) = $4,100
So, total expected generation cost:
- (0.8 × 2200) + (0.2 × 4100) = $2,580
Similarly, without the Baseload running:
- High-wind scenarios cost $4,000
- More importantly, low-wind scenarios requires 10 MW of load shedding, increasing he total cost to $107,000
- So, the total expected generation cost without the baseload running is $24,600

New Zealand: Baseload plant’s private economic incentives

Remember, firms just bid a marginal cost, not a 3-part cost
In the high-wind scenario:
- It sells 40 MW at a real-time price of $30/MWH
- Total Revenue: 30 × 40 = $1,200
- Total Cost: 30 × 40 + 1000 = $2,200
- Total Profit -$1,000
In the low-wind scenario:
- It sells 70 MW at a real-time price of $100/mwh
- Total Revenue: 100 × 70 = $7,000
- Total Cost: 30 × 70 + 1000 = $3,100
- Total Profit: $3,900
So, total expected profit:
- 0.8 × -1000 + 0.2 × 3900 = -$20
As the expected profit is negative, the baseload plant will not commit

New Zealand: Risk transfer

McRae argues that the single settlement design systematically transfers risk from wind to baseload plants.
For Wind:
- Wind profits are stable: $1,800 in the high-wind scenario and $2,000 in the low-wind scenario
- This is due to the negative correlation between prices and quantities for the wind plant
- Since intermittent sources have volatile but largely correlated production, this problem will only become worse with the clean energy transition
For Baseload plants:
- Even though they are essential for satisfying demand at a reasonable cost, they absorb most of the financial volatility created
- With the clean energy transition, baseload plants face higher profit volatility, higher cost of capital, and hence weaker investment incentives exactly when the system needs additional backup capacity
- Australia has ageing Coal Plants. According to BNEF, 70% of Australia’s coal fleet could retire by 2035. This amplifies the harm of weaker investment incentives.

New Zealand: Multi-Settlement Markets

Multi-settlement markets create financially binding positions before the uncertainity is resolved
Generators submit offers to a day-ahead market. The system operator prepares a dispatch schedule for the following 24 hrs. A real-time balancing market handles any deviation from the day-ahead schedule.
Generators receive the day-ahead price for their scheduled generation, and then pay/receive the real time prices for deviations from the schedule.
Going back to the blackout that happened in 2021, one contributing factor there was that two thermal plants were not operating. One plant had not started in the morning because the forecasted prices in the pre-dispatch were too low to cover it’s startup costs.
Two Benefits:
- Multi-settlement markets use three-part offers that include start-up costs, no load costs and marginal operating costs. This enables two mechanisms:
  - Compensating generators with make-whole payments when day-ahead revenues are insufficient
  - Allowing System Operators to solve for the optimal 24-hour dispatch that minimizes expected generation costs subject to operating constraints
- Thermal Generators can sell expected output in the day-ahead market, securing revenue before demand uncertainty resolves, and hence are guaranteed to recover startup and operating costs.

New Zealand: Multi-Settlement Markets (Continued)

Let’s revisit under Multi-settlement markets

Table 1: Characteristics of generation plants
Plant	Capacity (MW)	Marginal Cost	Startup Cost ($)	Lead Time (Periods)
Wind (High)	60 (p = 0.8)	0	0	0
Wind (Low)	20 (p = 0.2)	0	0	0
Baseload	70	30	1000	1
Peaker	70	100	0	0

New Zealand: Multi-Settlement Markets (Continued)

Suppose in the day ahead market, the Wind Plant sells 40 MW and the Baseload plant sells 60 MW
The Day-Ahead price = Expected Real-Time price = 0.8 × 30 + 0.2 × 100 = $44
At this price, the day-ahead revenues are:
- Wind Plant: 44 × 40 = $1,760
- Baseload Plant: 44 × 60 = $2,640
Since the as-bid costs of the baseload plant are $2,800 (1,000 + 30 × 60), it will receive a make-whole payment of $160
- So the Baseload plant’s total day-ahead revenue will be: 2640 + 160 = $2,800

New Zealand: Multi-Settlement Markets (Continued)

In the high-wind scenario:
- The Wind Plant sells an additional 20 MW in real-time at $30
  - It’s total profit will be: (44 × 40) + (30 × 20) = $2,360
- The baseload plant buys back 20 MW at $30
  - Total Revenue: (44 × 60) + 160 - (30 × 20) = $2,200
  - Total Cost: 1000 + (30 × 40) = $2,200
  - Total Profit: $0
In the low-wind scenario:
- The Wind Plant buys back 20 MW at $100
  - It’s total profit will be: 44 × 40 - 100 × 20 = -$240
- The baseload plants sells an additional 10 MW in real time at $100
  - Total Revenue: 44 × 60 + 160 + 100 × 10 = $3,800
  - Total Cost: 1000 + (30 × 70) = $3,100
  - Total Profit: $700

New Zealand: Multi-Settlement Markets (Why does it work?)

The economic logic at play here is that it assigns a price premium reflecting the value of dispatchability.
When wind sells less in the day ahead market than it produces:
- Real-time prices fall below day-ahead prices
- Wind plants can sell excess output at lower prices
- Baseload plants must buy-back their day-ahead commitments at lower prices
When wind sells more in the day ahead market than it produces:
- Real-time prices exceed day-ahead prices
- Wind plants must buy-back their day ahead commitments at higher prices
- Baseload plants can sell additional output at higher prices
Not only does this strategically benefit baseload plants, it also incentivizes Wind plants to make more accurate predictions of their output
Impediments to this accuracy can be strategic or technical

Conclusion

In Conclusion,
- Australia should:
  - Implement locational marginal pricing
  - Make loss factors dynamic
  - Make retail demand price-responsive
- New Zealand should:
  - Implement multi-settlement markets
Bonus chocolate question! Maybe all of this would have been avoided if the First Chair of the State Electricity Commission of Victoria had implemented these reforms. Who was the first chair?

Miscellaneous (if time permits)

McRae’s argument on market power is similar to David M. Newbery’s “Contract Market” argument in Power Markets and Market Power
Both papers argue for multi-period optimization in place of interval-by-interval optimization
Day ahead markets also induce pro-competitive behavior since they decrease the size of the spot market
Combinatorial bids can also help