Research And Realization Of Partial Discharge On-line Detection Technology For Power Distribution Equipment

With the expansion of the power grid,the traditional regular off-line maintenance mode can no longer meet the increasing demand for power reliability.Partial discharge,as an early phenomenon of insulation damage of electrical equipment,is an important basis for judging the insulation status.Based on the study of the on-line PD detection,a multi-function portable partial discharge on-line detection instrument is designed.Its multi-channel signal acquisition function can meet the requirements of detection for different equipment.Its diversified data analysis function provides users with more comprehensive and specific information.The BP neural network equipped with the function of PD type recognition is of great significance for reducing the cost of expert recognition and improving the accuracy of recognition.In this paper,the research background and status of on-line detection and pattern recognition of PD are described,the deficiencies of most PD detection instruments in the realization of pattern recognition are explained,and the limitations of existing research on PD pattern recognition in the field of on-line detection are pointed out.So,the requirement of a portable multi-function PD on-line detection instrument combined with pattern recognition technology is presented.The mainstream on-line PD detection methods for power distribution equipment are analyzed and compared.Based on this,the signal acquisition methods of the designed instrument including TEV,ultrasonic and high-frequency current are determined.The principle and selection of sensors for PD signals are studied,and the corresponding signal conditioning circuits are designed.Based on the mainstream function and structure of portable PD detection instruments,hardware design and debugging which include power management,memory components,communication interfaces,and various peripherals have been completed.The design of FPGA and processing system is completed on the ZYNQ embedded system platform,and the functions of signal intensity,pulse number,pulse waveform real-time display,storage of PRPD spectrum,replay of sound and so on are realized.Finally,the technology of feature extraction and pattern recognition of PD signals are studied.With the aim of simplification and real-time,the algorithm of simplified low dimension hybrid feature extraction is designed.And then this algorithm is transplanted to the designed PD detection instrument together with BP neural network,which realizes the fast and real-time intelligent classification of PD type.The accuracy of recognition is relatively high,which proves the validity and rationality of the algorithm used in the local power line detection system.