Bridging theory and practice: Efficiently solving the unit commitment problem in production cost models

High-fidelity Production Cost Models are tools for scheduling the operations of power systems, but their computational complexity often necessitates trade-offs in model detail or solution quality. We argue that there are tangible opportunities for resolving this trade-off, or at least mitigating it. To explore this hypothesis, we contribute a holistic approach that simultaneously addresses both model building and solution process. Our model building relies on recent advances in the formulation of Unit Commitment Problems, while our solution processes is based on two novel heuristics (iterative rounding and column generation) that exploit specific attributes of the unit commitment problem. Extensive computational experiments carried out on country-scale grids (Laos, Cambodia, and Thailand) show that iterative rounding achieves substantial speed-ups (up to 10x) while maintaining an average optimality gap of no more than 1% with respect to our benchmark solver (Gurobi). Notably, this heuristic also scales very well with the problem size. Column generation is less successful in terms of runtime, but demonstrate potential for instances characterized by high penetration of renewables. Both heuristics are solver independent, allowing a seamless integration with virtually any mathematical-programming solver. The diverse characteristics of the power systems we experimented on further indicates the generalizability of the proposed heuristics, highlighting their potential to improve the computational efficiency of production cost modeling.