气隙放电电压的大气条件灰联度分析及预测  

Gray Correlation Analysis of Atmospheric Conditions and Prediction of Air Gap Discharge Voltage

作　　者：牛海清[1] 许佳[1] 吴炬卓 余佳[1] NIU Hai-qing XU Jia WU Ju-zhuo YU Jia(School of Electric Power, South China University of Technology, Guangzhou 510640, Guangdong, China Zhuhai Power Supply Bureau, Zhuhai 519000, Guangdong, China)

机构地区：[1]华南理工大学电力学院,广东广州510640 [2]珠海市供电局,广东珠海519000

出　　处：《华南理工大学学报（自然科学版）》2017年第7期48-54,共7页Journal of South China University of Technology(Natural Science Edition)

基　　金：国家高技术研究发展计划(863计划)项目(2015AA050201)~~

摘　　要：为了研究大气条件参数对空气间隙放电电压的影响程度,使用放置在自然环境中的球-球电级全自动放电监测装置实时监测和记录的放电电压和大气条件参数数据,建立灰色关联度的计算模型,并通过计算得到各大气条件参数对放电电压的灰色关联度,结果表明,大气条件参数按灰色关联度大小(从大到小)的排序依次为气压、温度、风速、相对湿度、照度。以大气条件参数为输入,使用Chebyshev神经网络对放电电压进行预测,取得比BP神经网络更好的预测结果.根据大气条件参数的排序,分别取前两者(气压、温度)、前三者(气压、温度、风速)、前四者(气压、温度、风速、相对湿度)作为Chebyshev神经网络的输入,对放电电压进行预测.预测结果表明,随着输入个数的减少,预测的平均相对误差和最大相对误差变化很小.In order to discover the impact of atmospheric condition parameters on air gap discharge voltage, a auto-matic discharge monitoring device with ball-ball electrode was used to monitor and record the discharge voltage and atmospheric condition parameters in natural environment, and a calculation model of gray correlation was estab-lished ,by which the gray correlations between atmospheric condition parameters and discharge voltage were ob-tained ,finding that the gray correlations of atmospheric condition parameters are indicative of the following o rde r： pressure 〉 temperature 〉 wind speed 〉 relative humidity 〉 illumination. Then, by taking the atmospheric condition parameters as inputs, Chebyshev neural network was used to predict the discharge voltage, with better prediction results being obtained in comparison with BP neural network. Finally, according to the sort of atmospheric condi-tion parameters, the first two （ pressure and temperature） , the first three （pres s ure, temperature and wind speed） and the first four （p ressure, temperature, wind speed and relative humidity） parameters were respectively taken as the inputs of Chebyshev neural network to predict the discharge voltage. The results show that , with the reduction of the number of inputs, the average relative error and maximum relative error of the predicted values both have little change.

关 键 词：大气条件 空气间隙 放电电压 灰色关联度 CHEBYSHEV神经网络 

分 类 号：TM83[电气工程—高电压与绝缘技术]