A decomposition and coordination approach for large-scale security constrained unit commitment problems with combined cycle units

MISO faces one of the most challenging Day-ahead Security Constrained Unit Commitment (SCUC) problems in view of its large number of units (including many virtual transactions) and complicated transmission capacity constraints. When combined cycle units with complicated state transitions are present, branch-and-cut shows poor performance in terms of solution quality and solving times. In this paper, a synergistic combination of Surrogate Augmented Lagrangian Relaxation and branch-and-cut is presented. With system coupling constraints relaxed and quadratic penalties on constraint violations added, the augmented Lagrangian is innovatively linearized and decomposed into subproblems to be solved by branch-and-cut. Subproblem solutions are then effectively coordinated based on Surrogate subgradients with much reduced computation requirements and multiplier zigzagging, and accelerated reduction of constraint violations. Numerical testing of a complicated MISO SCUC case demonstrates that our method generates near-optimal solutions in short solving times and significantly outperforms branch-and-cut.