Abstract:Aiming at the abnormal state of transmission line in the long-term operation and maintenance process, the manual periodic line inspection method is mainly used for troubleshooting, and the hidden danger cannot be predicted efficiently and accurately. This paper presents a method for pre-discharge identification of typical hidden dangers in transmission lines based on BP neural network. Firstly, a typical test platform for transmission lines is established using Key Laboratory of Ice Prevention and Disaster Reducing of China Southern Power Grid in Plum Blossom Hill. The pre-discharge pulse current waveform data of hidden dangers at the transmission line under typical tree barrier and polluted insulator are obtained. Then the discharge quantity, discharge times and phase are extracted. These parameters are extracted from these data to obtain a feature parameter database, and these feature quantities are brought into the back-propagation(BP) neural network classifier for training the model of hidden dangers. Finally, the model test results show that BP neural network algorithm can effectively identify typical types of hidden dangers in transmission lines and the recognition accuracy rate is above 92%, which provides a reference for the diagnosis of hidden dangers in transmission lines.