Partial discharge diagnosis algorithm for multi-source ultrasound detection based on time series integration

The method based on ultrasound detection is commonly used in the scenes of on-line monitoring and live detection for partial discharge of power equipment. However, due to the different detection time, detection type and monitoring mode, the ultrasound detection data is isolated. Based on the condition, a partial discharge diagnosis algorithm for multi-source ultrasound detection based on time series integration is designed. Based on different types of ultrasonic amplitude maps, ultrasonic waveform maps, ultrasonic phase maps, ultrasonic flight maps, and ultrasonic audio sound maps of power equipment, the algorithm designs different single-map diagnosis algorithms according to different features of ultrasound maps. The time series integration is acted for diagnosis results of multiple maps in the same source of partial discharge, and an integration learning algorithm based on Adaboost is designed. Through training and learning, the source type of the partial discharge is obtained to achieve the effective diagnosis of the partial discharge signal. Compared with the traditional diagnosis algorithm for partial discharge based on single type and single map, this method with strong robustness can adapt to different partial discharge recognition scenes based on ultrasound detection, and effectively improves the accurate recognition for PD type of power equipment.