Research On Partial Discharge Detection And Pattern Recognition Of 10kV Switchgear

The 10kV switchgear is used as a complete set of electrical equipment in the power distribution system.Its safe and stable operation is the basis for ensuring reliable power supply of the power distribution system.Partial Discharge(PD)has always been the main cause of insulation degradation of the internal equipment of the switchgear,which affects the safe and stable operation of the switchgear.Transient Earth Voltage(TEV)is a more effective on-line detection method for partial discharge of switchgear.It has been widely used in maintenance and operation.However,it can only qualitatively analyze the degree of hazard,there is a lack of effective quantification of partial discharge,localization of TEV sensor optimal detection position,and pattern recognition in field application.This paper studies this,the main achievements are as follows:Firstly,according to the actual switchgear insulation defects,four typical switchgear partial discharge models of corona discharge,surface discharge,cavity discharge inside solid insulation and free particles discharge are designed and fabricated,and the experimental system is built.The relationship between the amplitude of the TEV sensor and the amount of discharge are studied with different partial discharge models respectively,the results show a positive proportional relationship.Then,the precision modeling of the switchgear is carried out according to the structural dimensions and material parameters,the electromagnetic wave propagation characteristics are simulated and analyzed,and the propagation law is obtained.In the case of four different discharges inside the switchgear,the influence of the TV signal volatility and insulation distance on different detection positions on the surface of the switchgear are analyzed comprehensively,and the optimal TEV sensor detection position is determined and verified by experiments.The results show that the simulation results are consistent with the experimental results.Then,a total of 5 6-dimensional characteristic parameters such as statistical characteristic parameters,gray image matrix characteristic parameters,time domain characteristic parameters and frequency domain characteristic parameters are extracted from the typical partial discharge models.The neural network algorithm is used to verify that the combined characteristic parameters are higher than the single characteristic parameters,but the recognition time is too long.Compared with PCA and local linear embedding dimension reduction algorithm(LLE),LLE is applied to reduce the dimension of the combined characteristic parameters,and the actual effect of the dimension reduction algorithm is verified by clustering recognition algorithm.Finally,a multi-classification support vector machine(MSVM)with strong generalization ability is selected as the pattern recognition algorithm,and its parameters are optimized by multi-group genetic algorithm(MPGA).The influence of different parameters of LLE reduction algorithn on the recognition rate of MPGA-MSVM algorithm is studied,which determine the optimal combination of parameters.The feasibility of the algorithm is further verified by comparing the recognition rates of PCA and LLE dimensionality reduction algorithms.Compared with BP,MSVM,MRVM and other classification algorithms,it is verified that the recognition effect of MPGA-MSVM algorithm is the best.The MPGA-MSVM algorithm trained by the typical model can identify the actual switchgear fault,and it can give the probability ratio of the type of discharge.It is proved that the MPGA-MSVM algorithm with LLE reduction algorithm has reliable reliability for partial discharge pattern recognition of the switchgear.The recognition accuracy can complete the recognition task better.