Solving stochastic hydrothermal unit commitment with a new primal recovery technique based on Lagrangian solutions

Abstract The high penetration of intermittent renewable generation has prompted the development Stochastic Hydrothermal Unit Commitment (SHUC) models, which are more difficult to be solved than their thermal-based counterparts due hydro constraints and inflow uncertainties. This work presents a SHUC model applied in centralized cost-based dispatch. is represented by two-stage stochastic model, formulated as large-scale mixed-binary linear programming problem. solution strategy divided into two steps. first step Lagrangian Relaxation (LR) approach, solve dual problem generate lower bound for SHUC. second given Primal Recovery where we use LR with heuristics based on Benders’ Decomposition obtain primal-feasible solution. Both steps benefit from each other, exchanging information over iterative process. We assess our approach terms quality solutions running times space scenario decompositions. computational instances various power systems, considering different configuration plants (capacity number units). results show advantage primal recovery technique compared solving via MILP solver. true already deterministic case, grows problem’s size (number and/or scenarios) does. decomposition provides better solutions, although one bounds, but main idea encourage researchers explore decompositions other relevant problems.