基于AMPSO和GRNN的绝缘子污秽度预测模型  

Prediction Model for Insulator Contamination Degree Based on AMPSO and GRNN

作　　者：李哲[1] 姜昀芃 卢明[1] 华奎 刘善峰 段晓红 

机构地区：[1]国网河南省电力公司电力科学研究院,河南郑州450052 [2]华中科技大学电气与电子工程学院,湖北武汉430074 [3]武汉云兆信息技术有限公司,湖北武汉430074

出　　处：《广东电力》2017年第7期76-82,共7页Guangdong Electric Power

基　　金：国家电网公司重大基础前瞻科技项目(SG20141187)

摘　　要：绝缘子表面沉积的污秽物是造成污闪事故的主要原因,而污闪事故对电力系统的安全稳定运行构成较大威胁,因此,需要定期对输电线路绝缘子进行检修和清扫作业以避免污闪事故发生。针对此,利用广义回归神经网络(general regression neural network,GRNN)来预测绝缘子表面等值附盐密度(equivalent salt deposit density,ESDD),但GRNN中平滑因子的取值对网络的预测性能有较大影响,因而利用自适应变异粒子群(adaptive mutation particle swarm optimization,AMPSO)算法来选取GRNN的最佳平滑因子。进一步提出经AMPSO优化的GRNN对绝缘子污秽度的预测模型,预测结果表明,所提优化预测模型可以对绝缘子的ESDD数值进行有效预测,且预测误差相对较小。Contamination on the surface of insulator is the main cause for pollution flashover, and accidents of pollute flashover may threaten safe and stable operation of the power system. Thus, it requires regular maintenance and cleaning work to avoid pollution flashover accidents. Therefore, this paper uses generation regression neural network （GRNN） to predict equivalent salt deposit density （ESDD） on the surface of insulator. But the value of smooth factor in GRNN has great impact on prediction performance of the network, it tries to use adaptive mutation particle swarm optimization （AMPSO） to select optimal smooth factor of GRNN. Furthermore, it proposes a GRNN prediction model for contamination degree of the insulator based on AMPSO and prediction results indicate the proposed model can perform effective prediction on ESDD of the insulator and the prediction error is comparatively less.

关 键 词：绝缘子 等值附盐密度 预测模型 自适应变异粒子群 广义回归神经网络 

分 类 号：TM216[一般工业技术—材料科学与工程]