水电站发电优化调度的综合改进差分进化算法  

Comprehensive Modified Differential Evolution algorithm for optimal operation of the hydropower station

作　　者：钟平安[1] 张卫国[1,2] 张玉兰 赵云发[3] 

机构地区：[1]河海大学水文水资源学院,江苏南京210098 [2]宁波市水利水电规划设计研究院,浙江宁波315192 [3]三峡梯调通信中心,湖北宜昌443133

出　　处：《水利学报》2014年第10期1147-1155,共9页Journal of Hydraulic Engineering

基　　金：973计划项目(2013CB036406);国家自然科学基金资助项目(51179044);国家重大科学研究计划(973)项目(2010CB951102)

摘　　要：差分进化算法能够全局并行和高效搜索,在求解水电站发电优化调度问题中具备一定优势,但同时存在一些问题。本文针对差分进化算法初始种群代表性不足的问题,引入基于均匀设计的初始种群生成方式;针对差分进化算法两种常用变异模式侧重于不同搜索阶段的特点,引入了兼顾两种模式优点的"Rand-Best耦合模式的变异算子";针对算法局部搜索能力欠缺的问题,借鉴梯度寻优思想,构建了"梯度加速附加算子"。以三峡水电站为例,说明了综合改进差分进化算法UCG-MDE的有效性,该算法大幅提高了解的精度、稳定性和收敛率。With parallel and efficient global search ability,Differential Evolution（DE） algorithm has cer-tain advantages for solving the optimal operation problem of hydropower generating,but there still existsome problems. In order to improve the representation of the initial population,a way of generating initialpopulation based on uniform design is introduced. Considering two common mutation operators of DE fit dif-ferent search phase respectively,the＂coupling Rand-Best mode of mutation operator＂with the advantagesof both is provided. In view of the fact that DE lacks local searching ability in some sense,the＂addition-al gradient acceleration operator＂is proposed,inspired by the gradient optimization ideas in classical optimi-zation algorithm. The results of application in the Three Gorges hydropower station show that the Comprehen-sive Modified Differential Evolution（UCG-MDE）is effective,significantly improving the accuracy,stabilityand convergence rate of solutions.

关 键 词：水电站 优化调度 差分进化算法 均匀设计 变异模式 梯度加速 

分 类 号：TV697.1[水利工程—水利水电工程]