Computational Study of Security Constrained Economic Dispatch With Multi-Stage Rescheduling

We model post-contingency corrective actions in the security-constrained economic dispatch and consider multiple stages of rescheduling to meet different security constraints. The resulting linear program is not solvable by traditional LP methods due to its large size. We devise and implement a series of algorithmic enhancements based on the Benders' decomposition method to ameliorate the computational difficulty. These enhancements include reducing the number of subproblems, solving the LPs faster by using appropriate solver options, harnessing parallel computing and treating difficult contingencies separately by integrating an independent "feasibility checker" process in the algorithm. In addition, we propose a set of online measures to diagnose and correct infeasibility issues encountered in the solution process. The overall solution approach, coded directly in GAMS, is able to process the "N-1" contingency list in 10 min for all large network cases (e.g., the Polish 2383-bus case) available for experiments.