水电站发电优化调度遗传算法的改进

Improvement of genetic algorithm for its application to optimal operation of hydropower station

遗传算法在求解水库优化调度问题时存在两个问题:随机的初始种群生成方式难以保证个体在解空间均匀分布,导致求解结果不稳定;由于水库水量平衡等条件约束,使交叉、变异操作常常导致可行解变成不可行解。针对上述问题,引入了基于均匀设计的初始种群生成方式,保障了初始种群空间分布的均匀性;在交叉、变异操作中建立了个体可行性判断与处理方法,保障了在个体进化过程中的可行性。以三峡发电优化调度为例,验证了改进算法的有效性,与标准遗传算法相比改进遗传算法可以显著提高寻优性能。

The genetic algorithm（GA） has been confronted with difficulties when used to solve the problem of reservoir operation.That is,the randomly-generated population that is not evenly distributed in solution space causes numerical instability,and the random operators usually generate infeasible individuals that violate the constraints.This paper presents an improved GA in combination with optimization of reservoir operation.A uniformly designed initial population is brought in first for insuring the uniformity of individuals,and then a procedure of checking individuals＇ feasibility is added to the cross and mutation operators.Application of this improved procedure to the modeling of Three Gorges reservoir validated the model＇s performance,and the result shows that the new algorithm is better than a standard GA in accuracy and efficiency.