(434f) Heuristic MINLP Strategies for Industry-Scale Unit Commitment with AC Power Flow

Real-time optimal scheduling of power generation and transmission is a fundamental component of the electric power ecosystem in modern power grids. Generator unit commitment decisions are made in a rolling horizon manner by market operators using a direct current (DC) approximation of a power network of producers and consumers. Network opera- tors then modify the optimal DC schedule to ensure feasibility and satisfy reliability requirements in the alternating current (AC) grid. While this strategy has been successful, incorporating AC power flow equations in the power generation scheduling problem has the potential to reduce the need for post-solve modifications of the implemented schedule and improve the efficiency of generation and transmission decisions, saving millions of dollars in operating costs. However, optimizing AC power transmission simultaneously with generator commitment (AC unit commitment) requires solution of a large-scale mixed-integer nonlinear program (MINLP) and has not been demonstrated in the time limits within which these real- time decisions must be made. To investigate whether these challenging MINLPs can be solved for industry-scale networks within practical time limits and compare solution approaches on standard datasets, ARPA-E has developed the Grid Optimization Competition.