Improved multiple-objective dynamic programming model for reservoir operation optimization

Abstract

Reservoirs are usually designed and operated for multiple purposes, which makes the multiple-objective issue important in reservoir operation. Based on multiple-objective dynamic programming (MODP), this study proposes an improved multiple-objective DP (IMODP) algorithm for reservoir operation optimization, which can be used to solve multiple-objective optimization models regardless whether the curvatures of trade-offs among objectives are concave or not. MODP retains all the Pareto-optimal solutions through backward induction, resulting in the exponential increase of computational burden with the length of study horizon. To improve the computational efficiency, this study incorporates the ranking technique into MODP and proposes an efficient IMODP algorithm. We demonstrate the effectiveness of IMODP through a hypothetical test and a real-world case. The hypothetical test includes three cases in which the trade-offs between objectives are concave, convex, and neither concave nor convex. The results show that IMODP satisfactorily captures the Pareto front for all three cases. The real-world test focuses on hydropower and analyzes the trade-offs between total energy and firm energy for Danjiangkou Reservoir. IMODP efficiently identifies the Pareto-optimal solutions and the trade-offs among objectives.