标准差分进化算法在求解梯级水库长期调度问题时,随着解链长度的增加,算法求解性能下降,易于陷入局部最优,为此构建了双层交互的混合差分进化算法。该算法结合差分进化算法的全局搜索和混合蛙跳的局部挖掘性能,上层基于小生境排挤技术保证差分进化算法的种群多样性,下层基于族群独立进化与定期通信保证局部搜索的稳定性。在整体上,下层结构为上层提供更多质量较优、结构互异的优良解,双层结构之间通过信息交互完成搜索信息的更新。通过将改进算法与标准差分进化以及逐次逼近动态规划法对比,模拟结果表明改进算法在求解梯级水库联合优化问题时的有效性和优越性。
This paper develops an algorithm of double-tier interaction shuffled differential evolution (DISDE) for optimal operation of cascade reservoirs to avoid performance degradation and local trapping of a regular differential evolution algorithm. This new algorithm integrates the regular global search ability and the local search ability of shuffled frog leaping algorithm, and guarantees a diversity of particles on its upper layer quipped with the niche technology and a local searching ability on its lower layer quipped with independent population evolutionary and regular correspondence. In the overall functions, the lower layer provides high- quality solution of different structures to the layer upon it, and searching information updating is completed by information interaction between the double-tier structure. Comparison with the regular differential evolution algorithm and dynamic programming of successive approximation, shows the efficiency and superiority of the new algorithm in optimization of cascade reservoirs operation.