Research On De-noising And Feature Extraction Of Leakage Current Of Transmission Line Insulators

The leakage current method is a good way to monitor the states of the online insulators on the transmission line. The de-noising and the feature extracting of the leakage current have been studied by some domestic and foreign scholars. And the pattern recognition methods also have been used to calssify the leakage current and to forecast the states of the insualtors. While, the methods have some faults. And many problems don’t be solved. In order to make up the faults of the methods, this paper studies not only the de-noising and the feature extracting of the leakage current but also the better pattern recognition for the leakage current classification.When the wavelet transform is used to de-noise the leakage current of the insulators on the transmission line, the calculation method of the best decompositon level number is proposed, and the best wavelet basis and the best threshold are given in this paper. During the de-noising with the wavelet threshold method, the wavelet coefficient is truncated with the threshold. The energy of the leakage current concentrates on the low frequency compocents. Based on the above two facts, the criterion judging the best decomposition level number is proposed. Based on this criterion, the formula is proposed and proved, which is used to calcalate the best docomposition level number of the wavelet transform. According to the analysis for the leakage current, the best wavelet basis and the best threshold are given.In order to de-noise the leakage current of the insulators on the transmission line adaptively, the intrinsic time-scale decomposition (ITD) method is improved, and the de-noising method based on the improved ITD is proposed. The distance scale is introduced to solve the rotation mixing problem of the ITD. And the extension of the two ends is introduced to solve the end effect problem. When the improved ITD is used to de-noise the leakage current, the formula calculating the decomposition number is proposed. Based on the wavelet threshold and the special characteristics of the improved ITD, the threshold and the de-noising process are given.The time domain entropy is proposed as the new feature of the insulator leakage current on the transmission line in the time domain in this paper. When the sampling frequency and the length of the leakage current are constant, the numbers of the absolute values of the leakage current in the different ranges are different. So, the concept distribution density is proposed. The entropy can reflect the distribution of the variables. Based on this concept "distritution density" and the characteristics of the entropy, the calculation process of the time doman entopy is given, which is proposed as the new method to extract the time domain feature of the leakage current.The frequency domain entropy is proposed as the new feature of the insulator leakage current on the transmission line in the frequency domain in this paper. The leakage current have obvious characteristics not only on their fundamental and the harmonic but also on their other frequency components. In order to reflect the change of the different frequency components of the field leakage current adaptively, the calcalation process of the frequency domain entropy is given, which is proposed as the method to extract the frequency domain feature of the leakage current.The multiclass relevance vector machine (m-RVM) is proposed as the mothod to recognize the insulator leakage current. The input vector, the output vector and the kernel function of the m-RVM are given. The process of recognizing the insulator leakage current is given.