Line Partial Discharge Monitoring And Fault Diagnosis Of Power Cable

When the power cable insulation slightly due to various reasons, it will have a partial discharge phenomenon. Partial discharge signals produced by its testing and to identify the type of discharge, the insulation of power cables to make judgments, to provide a theoretical basis for the system to repair or replace the power cable.At present, the partial discharge monitoring and identification system is mostly based on a single artificial defect model design, and when there are multiple Bureau put the source or sources of interference will cause the system to the testing environment may not be able to make accurate judgments on the discharge mode. The key to solving this problem is how to discharge signals and interference signals of the different types of mutual points from the modified fuzzy C-means clustering algorithm based on this idea in this article combined with support vector machine algorithm to establish a set of partial discharge online testing and Fault Diagnosis System.Discharge signals in time domain and frequency domain analysis, and the introduction of the equivalent width and the concept of equivalent bandwidth, the extraction of the discharge pulse signal when the equivalent width and equivalent bandwidth as classification characteristics. And the equivalent frequency formula to expand, if more information is needed to describe the characteristics of the discharge pulse characteristics to describe the information of the discharge pulse can be extracted by expanding the formula.Fuzzy C-Means clustering algorithm is a widely used cluster analysis algorithm, This algorithm is implemented through the iterative hill climbing method, so the algorithm is easy to fall into the initialization is particularly sensitive to local minima and not global optimal solution. Dynamic update the cluster centers to solve this problem this ingenious solution to this problem. The same time, this algorithm needs a pre-given number of clusters, but our pre-given number of clusters is inappropriate, will result in the real structure of the data is destroyed. To solve this problem, this paper presents a new cluster validity function is used to determine the optimal number of clusters, the system will be based on this function to output the optimal number of clusters and the optimal classification results. In pattern recognition, this Q using traditional PRPD statistics operator constitutes a discharge fingerprint feature vector, combined with support vector machine algorithm to make a judgment on the discharge mode. The support vector machine is a new kind of intelligent algorithm in pattern recognition, and the algorithm has many advantages. Out through the analysis of test results, the amount of equivalent time-frequency extraction for classification features and improved FCM clustering algorithm for accurate classification of the discharge pulse sequence, and support vector machine on the discharge mode recognition, the recognition results can be achieved to a satisfactory degree. This is for the development of experimental and theoretical basis for partial discharge recognition system based on a single defect model.