Learning-Assisted Variables Reduction Method for Large-Scale MILP Unit Commitment

The security-constrained unit commitment (SCUC) challenge is solved repeatedly several times every day, for operations in a limited time. Typical mixed-integer linear programming (MILP) formulations are intertemporal nature and have complex discrete solution spaces that exponentially increase with system size. Improvements the SCUC formulation and/or method yield faster hold immense economic value, as less time can be spent finding best-known solution. Most machine learning (ML) methods literature either provide warm start or convert MILP-SCUC to continuous formulation, possibly leading sub-optimality infeasibility. In this paper, we propose novel ML-based variables reduction accurately determines optimal schedule subset of trusted generators, shrinking dramatically reducing search space. ML indicators sets created shrink model, improvement quality. Test results on IEEE systems 14, 118, 300 busses, Ontario system, Polish 2383 3012 busses report significant reductions range 48% 98%. This promising tool operators solve lower optimality gap limited-time operation, benefits.