Hydropower generation management under uncertainty via scenario analysis and parallel computation

Abstract

We present a modeling framework for the robust solution of hydroelectric power management problems with uncertainty in the values of the water inflows and outflows. A deterministic treatment of the problem provides unsatisfactory results, except for very short time horizons. We describe a model based on scenario analysis that allows a satisfactory treatment of uncertainty in the model data for medium and long-term planning problems. Our approach results in a huge model with a network submodel per scenario plus coupling constraints. The size of the problem and the structure of the constraints are adequate for the use of decomposition techniques and parallel computation tools. We present computational results for both sequential and parallel implementation versions of the codes, running on a cluster of workstations. The codes have been tested on data obtained from the reservoir network of Iberdrola, a power utility owning 50% of the total installed hydroelectric capacity of Spain, and generating 40% of the total energy demand.