Author(s): Fangfang Li; Jiahua Wei; Xudong Fu; Xinyu Wan
Linked Author(s): Jiahua Wei
Keywords: Genetic algorithm; Incremental dynamic programming; Premature convergence; High di-mensional; Initialization
Abstract: Genetic algorithm (GA) has been widely applied to the optimization of reservoir operations due to its robustness in solving highly non-linear, non-convex problems. However, this application meets the difficulties such as premature convergence and convergence speed, which arise from high dimensional variable vector and large feasible domain of each variable. This paper is intended to propose an initialization policy for GA in the optimization using the Incremental Dynamic Programming (IDP). The optimum of IDP is obtained at first as the preliminary optimal solution. A neighbourhood of the IDP optimum is then defined as the feasible domain where the initial population of GA is generated. The IDP-GA hybrid approach was applied to daily optimization of the Three Gorges-Gezhouba cascade hydropower stations for the period of a year. It is shown that the hybrid approach is capable of effectively increasing the likelihood of reaching the global optimum. It would be a potential approach to dealing with high-dimensional optimization problems of reservoir operations.
Year: 2011