Research On Partial Discharge Characteristics And Pattern Recognition Of Composite Insulation Defects In High Voltage Generator Stator Bars

As the important equipment that undertakes the task of power output in the power system,the safety and operation stability of large-scale HV generators can directly affect the reliability of the power system.The faults of HV generators are mostly caused by the aging of the main wall insulation on the stator windings.Various types of insulation defects can occur at the same time due to aged main wall insulation and cause partial discharge in the stator.The discharge signal varies due to different insulation defects.Through the monitoring and analysis of the partial discharge,the timely and effective evaluation of the main wall insulation performance can be achieved,which is of great significance to the safe operation of the HV generator.The stator bar of the 13.8k V generator is taken as the research object.The causes,hazards,and mechanisms of discharges in the stator windings of the HV generator were summarized and analyzed.Based on this,four generator stator bar models with different composite defects were made for partial discharge tests.A climate chamber,partial discharge test electrode structure and test circuit were designed and built for partial discharge tests under different temperature and humidity conditions.The adjustment steps of temperature and humidity and voltage boosting methods were established.Based on the established HV generator stator bar models with composite defects,the composite partial discharge pattern on the main wall insulation of the HV generator stator was studied and analyzed from its partial discharge tests under different voltage levels,temperatures,and humidity.The results show that the voltage level can significantly change the partial discharge scale and the shape of the discharge pattern.The partial discharge scale increases with the increase of the applied voltage,and the discharge scale of various discharge components increases to different degrees with the increase of the voltage.The increase of temperature can promote the progress of corona discharge,but high temperature can bring more negative effects on corona discharge.The negative effect of temperature on bar-to-bar discharge in slot exit,slot discharge and surface discharge in slot exit is less than its positive effect.Humidity has a greater negative effect on corona discharge,bar-to-bar discharge in slot exit and surface discharge in slot exit,while slot discharge is less affected by humidity.In order to intelligently identify the partial discharge types of composite defects in HV generator stator bars,a graph convolutional network was built based on graph theory and graph structure.The partial discharge map was converted into the graph structure signal,and the recognition performance of the graph convolutional network under different hyperparameters was deeply studied.Then,the recognition results of the BP neural network,convolutional neural network and graph convolutional network for partial discharge types are compared.The results show that the graph convolutional neural network has a higher recognition rate and can provide a broader approach for intelligent recognition of partial discharge types of composite defects in HV generator stator bars.