摘要
总结了粒子群(PSO)算法的一些改进方法;分析并指出了PSO算法收敛困难的关键原因;提出了局优分支优化技术。该技术由5要素组成:①局部最优区域的确定;②局部最优区域的闭锁;③局部最优区域的深度搜索;④全局搜索的粒子补充;⑤迭代终止判据。还结合电网规划的特点提出了采用启发式逐步倒推模型对局部最优子群进行深度搜索的技术。在电网规划中的应用表明,该2项技术克服了PSO算法的收敛困难,提高了PSO算法的搜索效率,保证了PSO算法的全局搜索性能和局部搜索性能。同时,也为其它算法提供了新的优化思路。
Summarized some improved methods of Particle Swarm Optimization （PSO）, analyzed its critical reasons of convergence dilemma, then put forward local best embranchment optimization method, which consists of five elements： determination of local optimum; lock of local optimum area; embranchment search in local optimum area; particles recruitment in main search process and termination criterion. And a local depth optimization method based on retrogress mode of constructive heuristic algorithm according to the characters of transmission network expansion planning is also put forward. Numerical simulation results demonstrate these two methods overcome the convergence dilemma, improved the search efficient and ensured its local and global search ability simultaneously. They also provide new ideas for other optimization methods.
出处
《中国电机工程学报》
EI
CSCD
北大核心
2005年第23期12-18,共7页
Proceedings of the CSEE
基金
国家自然科学基金项目(50177017)
上海市重点科技攻关计划项目(041612012)
国家电网公司课题项目SGZL[2004]151~~
关键词
电力系统
电网规划
粒子群算法
启发式方法
局部最优解
全局最优解
分支优化
Power system
Power transmission network planning
Particle swarm optimization
Heuristic method
Local best solution
Global best solution
Embranchment optimization
