Interior-Point Method for Reservoir Operation with Stochastic Inflows

Abstract

A new method is proposed for long-term reservoir operation planning with stochastic inflows. In particular, the problem is formulated as a two-stage stochastic linear program with simple recourse. The stochastic inflows are approximated by multiple inflow scenarios, leading to a very large deterministic model which is hard to solve using conventional optimization methods. This paper presents an efficient interior-point optimization algorithm for solving the resulting deterministic problem. It is also shown how exploiting the problem structure enhances the performance of the algorithm. Application to regulation of the Great Lakes system shows that the proposed approach can handle the stochasticity of the inflows as well as the nonlinearity of the operating conditions in a real-world reservoir system.