基于神经网络的局部放电脉冲特征参数选择方法  

Method for the Characteristic Parameter Selection of PD Pulse Based on Neutral Network

作　　者：王晓蓉[1] 杨敏中[1] 董明[1] 严璋[1] 

机构地区：[1]西安交通大学电气工程学院,710049

出　　处：《电工技术学报》2002年第3期72-76,共5页Transactions of China Electrotechnical Society

摘　　要：提出了基于神经网络的特征参数选择方法 ,并利用该方法对所给出的表征局部放电脉冲特性的特征参数进行了选择。模式识别的结果表明 ,所提出的特征参数的选择方法不仅是有效的 ,而且是必要的。使用该方法时 ,对样本进行预处理是必须的 ,因它直接影响最终的结果 ,甚至会使该方法失效。本文给出了针对局部放电脉冲的去噪方法 ,并提出对样本数据进行处理时 ,宜综合应用两种归一化方式 。A method for the characteristic parameter selection based on neutral network was given in this paper,and with this method,the more sensitive and effective characteristic parameters were selected from the parameters which characterize the partial discharge(PD)pulse The results of pattern recognition show that the proposed method is not only valid but also necessary Moreover,the preprocessing of the samples was essential to use the method,it will influence directly the result of parameter selection and go so far as to cause the method ineffective As to PD pulse,the method of denoise is introduced also,and a notion to process the sample data is put forward apply synthetically with two ways of normalization,i e partial normalization and global normalization

关 键 词：局部放电 特征参数选择 神经网络 归一化 

分 类 号：R338.8[医药卫生—人体生理学]