Learning to Fix: Optimisation-Aware Machine Learning for Accelerated Unit Commitment
Unit Commitment is a computationally demanding mixed-integer linear optimisation problem requiring many binary commitment decisions across a scheduling horizon. To reduce this computational burden, machine learning approaches can predict a subset of these decisions, thereby shrinking the search space explored by the optimisation solver. Existing confidence-based approaches determine which variables to fix based on user-defined probability thresholds, which are agnostic to the downstream optimisation problem. We address this limitation by introducing an optimisation-aware framework that yields generator-specific confidence thresholds based on the impact of fixing errors on Unit Commitment solution quality. This approach won first place in the 2025 EPRI AI-ccelerating Unit Commitment competition. The proposed framework achieves a mean optimality gap below 0.5% while also delivering an average speed-up of more than 20x. It therefore provides a general mechanism for translating probabilistic predictions into controlled reductions of mixed-integer search spaces, directly linking learning decisions to their downstream computational and economic consequences.