基于BP-GA算法的染污支柱复合绝缘子伞裙结构优化  

Shed structure optimization of polluted post composite insulator based on BP-GA algorithm

作　　者：瞿王健 陆金桂[1] QU Wangjian;LU Jingui(College of Mechanical and Power Engineering,Nanjing Tech University,Nanjing 211800,China)

机构地区：[1]南京工业大学机械与动力工程学院

出　　处：《南京工业大学学报（自然科学版）》2019年第6期703-709,共7页Journal of Nanjing Tech University(Natural Science Edition)

摘　　要：为防止污闪事故的发生,提出了基于BP-GA算法的染污支柱复合绝缘子伞裙结构优化方法。基于优化拉丁超立方设计法选取大伞伸出、小伞伸出和伞间距作为伞裙结构设计变量,在多耦合物理场Maxwell软件中构造染污支柱复合绝缘子的电场计算模型,将针对不同伞裙结构求解出的沿面电场不均匀系数和最大沿面电场强度作为样本数据。利用BP神经网络对样本数据进行学习训练,建立电场特性近似模型。采用GA算法求解近似模型,在满足最大沿面电场强度作为性能约束的前提下,使得沿面电场不均匀系数最小。实例结果表明,沿面电场不均匀系数降低了23.16%,且最大沿面电场强度小于空气击穿场强,有效地提高了支柱复合绝缘子的污闪性能。In order to prevent the occurrence of pollution flashover accidents,a method of shed structure optimization of polluted post composite insulator based on back propagation-genetic algorithm(BP-GA)algorithm was proposed.The overhang of sheds and the spacing between sheds were selected as design variables based on the optimal latin hypercube design method.An electric field calculation model of polluted post composite insulator was established in the multi-coupling field analysis software Maxwell.The maximum surface electric field strength and the non-uniform coefficient of surface electric field corresponding to different shed structures were calculated as learning data.The back propagation(BP)neural network was used to learn and train the sample data,and the approximate model of electric field characteristics was set up.Under the condition that the maximum surface electric field strength was satisfied for the performance constraint,the genetic algorithm(GA)was used to solve the approximate model,and the minimum non-uniform coefficient of surface electric field was obtained.The optimization results showed that the minimum non-uniform coefficient of surface electric field was reduced by 26.16%,and the maximum surface electric field strength was reduced less than air breakdown electric field strength,which effectively improved the pollution flashover performance of post composite insulator.

关 键 词：染污绝缘子 伞裙优化 电场特性 神经网络 遗传算法 

分 类 号：TM216[一般工业技术—材料科学与工程]