While Economic Dispatch decides what power plants should do *right now*, Unit Commitment (UC) is the massive algorithm that decides what they should do *tomorrow*. Because it takes 12 hours and $50,000 in fuel just to warm up a massive coal or nuclear plant, grid operators must run the UC algorithm a day in advance to mathematically lock in exactly which massive turbines need to be physically turned on to meet tomorrow's peak demand.
Unit Commitment is a severely difficult Mixed-Integer Programming (MIP) problem. It involves 'binary variables' (0 = OFF, 1 = ON). You cannot turn a 1000MW coal plant on for just 1 hour. It has strict physical engineering constraints: a Minimum Up Time (if you turn it on, it must run for at least 24 hours to prevent thermal cracking of the boiler) and a Minimum Down Time. The SCUC (Security Constrained Unit Commitment) algorithm takes tomorrow's forecasted weather, solar output, and load, and crunches millions of permutations to find the absolute cheapest combination of plants to turn on (Commit) over a 24-hour horizon, while maintaining enough spinning reserve to survive a sudden fault.
Load and VRE Forecasting
At 10:00 AM, the grid operator ingests advanced AI weather models predicting exactly how much wind and solar (VRE) will be available for all 24 hours of the next day.
Bid Processing
Generators submit their complex parameters: 'It takes me 6 hours to start, costs $50k to start, and I must run for at least 12 hours.'
Day-Ahead Clearing
The SCUC supercomputer crunches the massive matrix. At 1:00 PM, it publishes the binding schedule, ordering specific massive thermal plants to begin their slow, 12-hour boiler ignition sequences.
- The Duck Curve: The massive influx of midday solar forces the algorithm to severely 'de-commit' thermal plants at noon, and then desperately 'commit' fast-ramping gas plants at 6 PM to cover the sunset, destroying the efficiency of the thermal fleet.
- Stochastic Uncertainty: Standard SCUC assumes the weather forecast is 100% perfect. Advanced grids are trying to shift to 'Stochastic Unit Commitment', which mathematically models the probability of the weather being wrong, but it requires 100x more supercomputing power.