Based On The Fuzzy Clustering Algorithm Of Partial Discharge On-line Monitoring System For The Design And Implementation

Partial discharge monitoring is the basis for the evaluation of high-voltageelectrical equipment insulation state, and it takes great significance for safe andeconomical operation of power system. By online monitoring of the high voltageelectrical equipment operating status, data acquisition and analysis the signal toindentify the mode of the partial discharge, it can detect the insulation defects and getthe discharge level, so as to prevent accidents. Therefore, the partial discharge patternrecognition has a very important value in both engineering and academic area. Currently,there are several partial discharge pattern recognition methods including fuzzyrecognition method, artificial neural network method, fractal method etc.Fuzzy clustering is an unsupervised classification method, with good intelligence itcan be applied to the analysis of partial discharge signals very well. In this paper, referto the existing partial discharge pattern recognition literature home and abroad, get alarge number of experimental data through simulation experiments of partial discharge.Studied the pattern recognition technique based on the fuzzy clustering, proposed animproved fuzzy identification method based on fuzzy c-means clustering algorithm. Themain subject of study as follows:1. Introduced the principle of the class merging algorithm based on fuzzy c-meansalgorithm and its implementation steps and characteristics in detail. The experimentsshow that the algorithm can be applied to the partial discharge signal analysis very well.2. Set up the partial discharge communications environment, achieved thecommunication between the system and the partial discharge test instrument, andreal-time display the signal.3. Established a database to save and manage the partial discharge signal by SQLserver, achieved the goal of unified maintenance, share with others and effectivemanagement of the data. Complete the intelligent analysis module with the algorithmadded.