基于惩罚的遗传蜂群算法在梯级调度优化中的应用

Application of Generic Bee Colony Algorithm based on Penalty in Cascade Dispatching Optimization

本文提出了一种结合了生物进化和群体智能思想的新型智能算法,并应用于水库群的梯级调度优化研究中。本算法以人工蜂群算法中群体协作的正反馈机制、个体分工的性态多样性思想、优良的全局搜索能力、并行计算性及较强的鲁棒性为基础,进行问题空间的全局寻优;在个体的局部寻优行为中,引入遗传算法的杂交和变异算子来优化侦查蜂路径,避免陷入早熟问题。同时针对梯级调度优化中常见的多维变量约束条件,借鉴模拟退火算法思想,在目标函数中构造的惩罚因子,使得带约束问题转化为了纯粹的优化问题。经实例验证,本算法具有普遍的梯级调度优化解决能力,并与传统的遗传算法及人工粒子群算法相比,具有更好的精度、收敛速度和寻优能力。

A new intelligent algorithm,combined of biological evolution and swarm intelligence,is proposed and applied to the cascade dispatching optimization in this paper.The algorithm optimizes in the globe space based on ABC algorithm which has positive feedback mechanism,division and cooperation mechanism,global search capability,parallel computing and good robustness.Crossover and mutation of GA are led into to avoid prematurity during local optimization.Penalty function of simulated annealing algorithm is also taken to simplify constraint handling.Compared with GA and PSO,this algorithm has better accuracy,convergence speed and optimization capability via an instance validating.