A Decomposition and Dynamic Programming Aggregation Method for the Optimal Water Allocation of Reservoirs in Series

Abstract

Research on water allocation of multiple reservoirs with the purpose of reducing water spills and improving the local runoff utilization is a matter of great concern in humid areas with uneven temporal and spatial distributions of water resources. An optimization model for a system of reservoirs in series is developed to minimize water shortages. Several constraints restrict the objective function, including available water, operation rules and water rights for replenishment of the reservoirs with water. The model features multiple dimensions with a single coupling constraint of the large-scale system. A decomposition and dynamic programming aggregation method (DDPA) is proposed; the subsystem models and the aggregation model are both solved with the classical one-dimensional dynamic programming. Compared with the conventional decomposition-coordination method, the proposed method is concise but reliable because it can directly use the results of subsystems to form the one-dimensional dynamic programming aggregation model, avoiding the iterative calculations according to the coordinating function. Compared with the meta-heuristic algorithms, the proposed method is more efficient because it is independent of any algorithm parameter. The proposed method may provide a new reference for solving similar multi-reservoir optimization models.