Improving Optimization Efficiency for Reservoir Operation Using a Search Space Reduction Method

Reservoir operation problems are challenging to efficiently optimize because of their high-dimensionality, stochasticity, and non-linearity. To alleviate the computational burden involved in large-scale and stringent constraint reservoir operation problems, we propose a novel search space reduction method (SSRM) that considers the available equality (e.g., water balance equation) and inequality (e.g., firm output) constraints. The SSRM can effectively narrow down the feasible search space of the decision variables prior to the main optimization process, thus improving the computational efficiency. Based on a hydropower reservoir operation model, we formulate the SSRM for a single reservoir and a multi-reservoir system, respectively. To validate the efficiency of the proposed SSRM, it is individually integrated into two representative optimization techniques: discrete dynamic programming (DDP) and the cuckoo search (CS) algorithm. We use these coupled methods to optimize two real-world operation problems of the Shuibuya reservoir and the Shuibuya-Geheyan-Gaobazhou cascade reservoirs in China. Our results show that: (1) the average computational time of SSRM-DDP is 1.81, 2.50, and 3.07 times less than that of DDP when decision variables are discretized into 50, 100, and 500 intervals, respectively; and (2) SSRM-CS outperforms CS in terms of its capability of finding near-optimal solutions, convergence speed, and stability of optimization results. The SSRM significantly improves the search efficiency of the optimization techniques and can be integrated into almost any optimization or simulation method. Therefore, the proposed method is useful when dealing with large-scale and complex reservoir operation problems in water resources planning and management.