用于局部放电模式识别的深度置信网络方法

Research of Partial Discharge Recognition Based on Deep Belief Nets

气体绝缘电器(gas insulated switchgear,GIS)内部绝缘缺陷产生的局部放电(partial discharge,PD),特征表现较复杂,分散性大,易受运行环境影响,而基于PD统计特征模式识别的传统方法,特征量选取主观性较强,且容易丢失部分特征信息,尤其对自由金属微粒类型缺陷识别率较低。因此,提出了一种基于深度置信网络(deep belief nets,DBN)的GIS设备内部PD模式识别方法,DBN能从数据中自主学习出高阶特征,避免了特征量选取的主观影响,能较好识别自由金属微粒类型缺陷,且识别用时远低于支持向量机(support vector machine,SVM)和BP神经网络(back propagation neural networks,BPNN)算法,作为对GIS设备PD模式识别的新方法具有一定的实用价值。

Partial discharge（PD） caused by GIS internal insulation defects has complicated and highly dispersed characteristics, susceptible to operation environment. Traditional methods of PD pattern recognition based on statistical characteristics are strongly subjective in feature extraction, easy to lose some characteristic information and have low recognition rate of free metal particle. Therefore, a PD recognition method based on deep belief nets（DBN） is proposed that can automatically capture PD spectrum high-order characteristics, avoiding subjective influence in feature extraction. Also, this method can recognize type flaws of free mental particle with recognition time far less than those of support vector machine（SVM） and back propagation neural networks（BPNN）, and therefore has practical values.