Study On Method And System Of Partial Discharge On-Line Monitoring In Switchgear Based On DSP

Partial discharge (PD) detection can effectively guarantee the safe and normaloperation of electric power equipment. At present, technologies and products of PDon-line monitoring are mainly used for transformer and GIS. However, they areseldom employed for switchgear, and regular testing and checking-type monitoravailable cannot promptly detect the breakdown of switchgear. As a result, by virtueof analyzing the research progress as well as developing trend of PD both at homeand abroad, this paper proposes the ultrasonic method and designs the PD monitoringsystem based on DSP, with the purpose to offer an real-time on-line detection on thePD of switchboard.In this paper, six discharge models are designed to simulate the partial discharge inswitchgear, including needle plate, suspended discharge, etc. Afterwards, the PDtesting system is constructed, and the ultrasonic data acquisition system is designedbased on data acquisition toolbox and GUIDE in MATLAB, which can acquire, store,and display the ultrasonic signal and lays an experimental basis for the off-lineanalysis of ultrasonic signal. On the basis of time-domain and frequency-domainanalyses of PD ultrasonic signals and the study on the influence of voltage, dischargedistance, suspended particles upon PD, two stable feature frequency bands areelicited.Due to the high sampling rate of ultrasonic signals, a regulate circuit of PD signalsis designed, comprising two filtering of ultrasonic signals, precise rectification as wellas integrating circuit. In this way, the integral values of ultrasonic signals arecollected, which decreases the requirement on sampling rate and increases the theprocessing rate of data. Accordingly, hardware system of PD on-line monitoring isbuilt with SEED-DPS2812M as the core board.The Support Vector Machine(SVM) is employed in pattern recognition of PD. Bymeans of utilizing the integral value of ultrasonic signal as characteristic features, andchoosing RBF kernel function to construct the binary classification model of Least Square SVM (LSSVM), PD is successfully identified. Then the principle of LSSVMis studied detailedly. CCS3.3works as a development tool in using C language totransplant LSSVM into DSP. And this algorithm plays a key role in designing theon-line monitoring software system, which consists of such programming modules asintegral value AD sampling, LSSVM pattern recognition, monitoring alarm of PD,etc.Experiments show that the on-line monitoring system based on DSP can stablyidentify PD and promptly warn beforehand. It sets the stage for developing theembedded products of PD on-line monitoring and has important practical significanceto ensure the normal operation of the switchgear.