Unit Commitment Model

This section outlines the two-stage, sequential optimisation framework we developed to forecast electricity prices and dispatch outcomes in the NEM. The approach balances long-term generation commitment decisions with short-term operational realities to enhance computational efficiency and market fidelity.

This approach is especially important for accurately modelling the high startup costs and slow ramping speeds of coal and CCGT generators. These characteristics significantly affect how these units bid into the market and how they are dispatched. As more renewables (primarily solar) comes online and increasingly supplies electricity during the middle of the day, prices during these periods are likely to fall. In response, coal and CCGT units, which currently operate as baseload generators, may find it uneconomical to stay online at their minimum stable load for hours at a time during these midday periods. Instead, they may choose to turn off and incur the startup costs of a warm or hot start later if it is more economical. This shift will have a growing impact on market prices and dispatch outcomes in the future.

Stage 1: Unit commitment

In the first stage, the model determines the optimal commitment status of thermal generation units across the forecast horizon. This is the initial run to optimise startup and shutdown for thermal technologies which have high startup costs and slow ramping. This first stage model run results in the commitments of black coal, brown coal, and CCGTs, which are then incorporated as an input to the second stage run. All other technology types are automatically set as committed and allowed to dispatch at optimal system cost in the second run.

Key constraints

• Minimum stable load (MSL): Binary variables flag operations below technical minimum loads
• Startup and inefficiency costs: Penalties applied when units operate below MSL to capture startup costs and inefficient performance
• Ramping limits: Physical ramp rate constraints on thermal generation
• Energy balance: Ensures market supply-demand equilibrium

Outputs

• Binary on/off decisions for each thermal unit
• These decisions constrain unit availability in the economic dispatch model (Stage 2)

Stage 2: Economic redispatch

Building on Stage 1, the second stage performs detailed economic dispatch, incorporating both unit availability and operational constraints.

Supply stack adjustments

• Commitment-driven availability: Thermal units committed in Stage 1 are available for dispatch. Conversely, if they are not committed, they are forced offline in this stage
• S-curve marginal cost adjustments: Applies volatility-aware adjustments to coal and CCGT short-run marginal costs to reflect realistic bidding behaviour

Enhanced operational constraints

• MSL enforcement: Units must operate at or above their technical minimum once committed
• Detailed ramping: Incorporates full ramping dynamics, including cost penalties for aggressive ramping
• Interconnector flows: Models cross-regional power transfers and associated losses
• Storage modelling: Includes pumped hydro and battery operations, accounting for round-trip efficiency and cycling limits

In summary

This modelling approach allows us to:

• Reflect realistic operations of coal and CCGT plants, and signals to turn on/off
• Capture dynamics (and to an extent, the costs) of MSL violations
• Emulate observed market bidding strategies with S-curve adjustments
• Reinforce physical realism in price-setting with ramping costs