以提升小波包系数熵为特征值的隐马尔科夫电缆局部放电识别

Partial discharge pattern recognition using hidden Markov models based on the entropy lifting wavelet coefficients

针对电缆局部放电检测,本文提出以提升小波包系数熵结合隐马尔科夫模型的识别方法。基于提升小波包与信息熵理论,提取放电信号的小波能量熵与系数熵作为特征量,将提取的特征向量输入隐马尔科夫模型进行训练,得到最优训练模型。在电缆本体上进行人工模拟缺陷,采用本文算法、传统小波系数熵、BP神经网络分别对不同放电模型产生的放电进行识别测试,并使用该方法对现场数据进行分析。结果表明:本文方法在识别准确率以及算法执行效率上,均优于传统小波以及BP神经网络。

For the detection of partial discharge in cables,this paper presents a recognition method based on lifting wavelet packet coefficient entropy and hidden Markov model.Based on the theory of lifting wavelet packet and information entropy,the wavelet energy spectrum entropy and coefficient entropy of discharge signal are extracted as eigenvalues.The extracted eigenvectors are input into the hidden Markov model for training,and the optimal training model is obtained.Artificial simulation of defects on cable body,discharges generated by different discharge models are identified and tested by using the proposed algorithm,traditional wavelet coefficient entropy and BP neural network respectively.The results show that the method is superior to the traditional wavelet and BP neural network in recognition accuracy and algorithm execution efficiency.