Abstract

ater stored in large reservoirs can support various economic activities, such as electricity generation and agriculture, that need to be efficiently coordinated. Such coordination is usually undertaken through optimization models that use hydro plants to minimize total electricity production cost subject to an array of prescribed constraints that ensure delivery of sufficient volumes of water to further sectors. In this context, we present a sub-optimal approach based on SDDP method to schedule power plants in a hydro-thermal system that presents non-convex irrigation constraints. We show applicability of this approach in the Chilean Central Interconnected System (SIC), where we accurately model current prescribed rules that coordinate water uses between electricity and agriculture sectors. We demonstrate that the presented SDDP-based approach determines feasible and near-optimal solutions with costs reasonably close to those optimally determined through alternative MILP formulation (that can properly capture non-convexity). We also found, however, that there may be issues associated with the levels of security of supply of the near-optimal solution since its stored water volumes tend to be lower than those associated with the optimum MILP solution.