Study On Partial Discharge On-Line Monitoring Technology Of High Voltage Cable

The aging breakdown of cable joint insulation is one of the main forms of cable failure.On-line monitoring of partial discharge in cable intermediate joints is of great significance for improving the reliability of cable operation.At present,most of the existing partial discharge monitoring technologies for cable joints are based on short-term monitoring data and threshold value tojudge insulation status,and the situation of missing and false alarm is serious.In order to improve the accuracy of insulation condition diagnosis,this dissertation designs an on-line monitoring system for partial discharge of cable intermediate joints,which can continuously collect and store partial discharge signals and extract discharge characteristic parameters.Then,BP neural network is used to evaluate and predict the insulation status of cable joints based on the development process of insulation aging and the long-term trend of characteristics.In this dissertation,the software design of PD signal acquisition system is completed in Lab VIEW,which includes acquisition,storage and interface display.The system sets the sampling rate of 20 MHz and stores 1 s data every 15 minutes.A storage mode based on producer-consumer cycle is designed.The upper computer software is used to control the oscilloscope and combine with HFCT and ultrasonic sensor to realize the real-time acquisition and storage of partial discharge data.Firstly,four kinds of physical models of typical insulation defects in cable intermediate joints are made in this dissertation.The discharge phase distribution characteristics of different discharge models were obtained.The discharge pulse of internal air gap discharge mainly concentrates on phase 30° to 80° and 200° to 260°.The discharge pulse of surface discharge mainly concentrates on 40° to 90°and 230° to 3000°.The quantity and times of discharge in negative half cycle of power frequency cycle are more than those of the positive half cycle.The burr discharge pulse is asymmetrical between positive and negative half cycle.The discharge pulse is mainly concentrated on 240° to 3200°.The discharge pulse of suspension discharge is mainly concentrated on 50° to 150° and 220° to 320°.The pulse density is the highest near 90° and 270°.Secondly,an accelerated aging test platform for 10 kV cable intermediate joints has been built in the laboratory.The accelerated aging test voltage is 20 kV and the cumulative aging time is more than 160 hours.During the aging process,the local discharge signals of the intermediate joints are collected continuously and the characteristic parameters are extracted.By comparing with the spectral characteristics of the defect model,the internal gap discharge of the cable intermediate joints is judged.A comparative study of the development process of electrical tree branches in partial discharge energy and insulation of middle and indirect joints shows that the three stages of PD energy change correspond well with the initiation period,the lag period and the growth period of insulation electrical tree branches.In the first stage,partial discharge occurs after nearly 50 hours of accelerated electrical aging.The amplitude of discharge energy and average discharge voltage increased slowly to 414.09 uJ and 131.36 mV respectively,and then entered the second stage.In the second stage,it lasts nearly 56 hours and the platform period is about 10 hours at the end of the stage.The eigenvalue amplitude is in a fluctuating state before platform period,which ranges from 16.82?J to 414.09 ?J and 77.32 mV to 131.36 mV.In the third stage,the average growth rate of total discharge energy and discharge pulse increases significantly compared with the first two stages and the amplitude of discharge energy and average discharge pulse finally increase to 1022.81?J and 207.14 mV.Finally,BP neural network models of different aging stages are established respectively,which takes discharge time,maximum discharge pulse,average discharge pulse and discharge energy in the characteristic parameters of historical data as input,discharge energy and average discharge pulse at future time as output.The BP neural network models of different aging stages are established respectively.After judging the aging stage,the untrained data in the samples are input into the corresponding model to predict the value about 60 days later.The relative error of prediction can be guaranteed to be less than 7%,which indicates that the model can evaluate and predict the insulation state of the cable intermediate joint before the insulation failure.