Algorithm Research Of Power Cable Fault Type Recognition Based On Oscillatory Voltage Partial Discharge Detection

Due to various reasons,the loss of power cable insulation occurs,the power cable will occur partial discharge phenomenon.The early insulation failure of the cable can be expressed by the partial discharge phenomenon,and the phenomenon of partial discharge will further promote the deterioration of the cable insulation.The insulation of the cable can be characterized by partial discharge.By detecting the partial discharge signal generated by the cable and identifying the type of partial discharge,the insulation condition of the cable can be known,and then the repair or replacement can be carried out in time to avoid the loss.Oscillating wave detection is a new and effective technique for partial discharge diagnosis in recent years.At present,the research and application of the method of oscillation wave detection is mainly in the two aspects of the detection and location of the cable defects,and the identification of the type of insulation defect under the oscillation wave detection is relatively short.Therefore,this paper summarizes the existing wave detection technology,recognition of partial discharge based on the combination of theory and experiment,carried out the research of Cable Partial Discharge defect recognition method of oscillating wave detection under,to accurately determine the cable insulation condition and defect type,prevent faults of cable lines,has an important application value for power system protection.This paper first introduces the generation of partial discharge and oscillation detection system,and then manually produced four types of XLPE cable joints typical defects: too long overlap of stress cone defect,metal cusp at high potential defect,air-gap in semi-conducting layer defect,cusp in semi-conducting layer defect.partial discharge signal data acquisition of various defects in the oscillating wave detection with the method,and the signal of the time domain and frequency domain analysis,mapping the data to the equivalent time-frequency map,Separating partial discharge signal and noise by using fuzzy clustering method,obtain clean partial discharge to be analyzed.For partial discharge of one-dimensional time signal,this paper adopts the partial discharge energy of wavelet packet decomposition and spectral features were identified by BP neural network based on recognition of good results,that the partial discharge energy of wavelet packet decomposition spectral features can effectively express the characteristics of the signal based on them;then on the current frontier deep learning method,which is introduced in this paper the deep belief network for recognition of partial discharge signals,the recognition effect is good,and the deep belief network can avoid the manual feature extraction steps,reduce artificial factors.For the random phenomenon of partial discharge,the PRPSA model into the principle of PRPD model,design the H_n (q,?) model of three dimension map of discharge signal;the discharging times of three-dimensional map characteristics and the statistical operator of threedimensional map characteristics is designed by using BP nerve identification experiment,results show that the statistical operator of three-dimensional map characteristics is better,it can reflect the cable partial discharge information good.