采用多样性引导粒子群算法的干式空心电抗器优化设计

摘要

According to the premature convergence of particle swarm optimization (PSO) algorithm, a diversity-guided attractive and repulsive particle swarm optimization (DGARPSO) algorithm is proposed for an optimal design of dry-type air-core reactor. Mutation is introduced into attractive and repulsive PSO (ARPSO) algorithm, which means that mutation is implemented to the particle positions in certain probability when the diversity of evolution population or personal best population gets less than the lower limitation. Thus the particles are promoted to fly away from the population aggregation position to effectively reduce the premature convergence of PSO algorithm in case of lower population diversity. The effects of uniform mutation, Gaussian mutation and Cauchy mutation on the optimization results are comparatively discussed. The simulation for a 50 kV ? A dry-type air-core reactor shows the better global search ability and performance of DGARPSO algorithm than GA algorithm,PSO algorithm and ARPSO algorithm.%针对粒子群优化(PSO)算法易于早熟收敛的问题,提出了采用多样性引导的吸引-排斥粒子群优化(DGARPSO)算法,并应用于干式空心电抗器的优化设计中.该算法在吸引-排斥粒子群优化(ARPSO)算法中引入变异操作,即当进化群体多样性或个体极值群体多样性小于下限值时,以一定概率对粒子的位置进行变异,从而使得粒子在群体多样性很低时飞离群体的聚集位置,有效减少了PSO算法的早熟收敛现象,同时还比较了均匀变异、高斯变异和柯西变异对优化结果的影响.对50 kV·A干式空心电抗器的仿真结果表明,DGARPSO算法提高了全局搜索能力,比GA算法、PSO算法和ARPSO算法具有更好的寻优性能.