Research On Statistical Characteristics And Pattern Recognition Of Partial Discharge Signals In Switch Cabinet

In the power system,the switch cabinet is an important intermediate equipment,which is responsible for receiving and distributing electric energy,and it plays a pivotal role as the "transfer station" of the power system.The normal operation of switch cabinet is the key to ensure the stability of power system operation.During long-term operation,electrical,thermal,chemical and other factors can cause partial discharge in the switch cabinet,which may lead to insulation failure.Therefore,the thesis investigates the statistical characteristics and classification of partial discharge signals in switch cabinet.The partial discharge experimental platform of high-voltage switch cabinet is built,and the ultrasonic sensor,TEV sensor and typical discharge models in the work site are designed.By simulating the partial discharge in the switch cabinet,the partial discharge signal data of ultrasonic and TEV under different voltage levels are obtained,which provides data support for the statistical analysis and classification of subsequent partial discharge data samples.The characteristics of partial discharge data samples are statistically analyzed,and the threshold parameters characterizing the severity of partial discharge are obtained.Firstly,based on the experimental data of ultrasonic and TEV partial discharge signals under different voltages,the amplitude information of ultrasonic and TEV partial discharge signal data obtained through the experiment is processed by using the pulse peak recognition method.Then,the test parameters of six probability distributions are compared,and the Weibull distribution and minimum extreme value distribution are selected to analyze the statistical characteristics of partial discharge ultrasonic and TEV signals of switch cabinet.Based on the analysis of the relationship between the amplitude of partial discharge signal,shape parameters and scale parameters,the threshold division criteria of partial discharge severity is proposed and the threshold parameters of partial discharge classification of switch cabinet are determined.Support vector machine and BP neural network are used to classify and identify the discharge severity of ultrasonic and TEV signals with the same data samples.Firstly,the characteristic quantities of ultrasonic and TEV discharge signals are extracted,and then the recognition rates of discharge severity from the two recognition models of support vector machine and BP neural network are compared.The results show that both support vector machine and BP neural network can effectively identify that there is no discharge,and the recognition rate of discharge severity of BP neural network is slightly better than that of support vector machine.The classification and recognition results to some degree can also verify the correctness of the threshold parameters of Weibull distribution.