After the automatic coverage of the distribution network, the massive signal data in the I area system is not effectively utilized and the switchgear in the distribution site has frequent and instantaneous flashovers and groundings before the fault and cannot be found in time. Distribution network defect sensing system based on waveform matching algorithm. Based on the instantaneous alarm signal and wave recording function of the distribution terminal (DTU) of the distribution line, this paper analyzes the data acquisition principle, waveform analysis and data processing, and finds and solves the existence of evolution through data mining. Equipment hidden dangers in the process; establish a distribution network defect perception model based on big data analysis and waveform matching algorithm, including data format model, convolutional neural network waveform learning model and risk assessment strategy. Finally, the feasibility of the design system is proved through the analysis of the distribution network defect sensing system example, which guides the on-site targeted partial discharge detection, and finds and solves the equipment hidden dangers in the evolution process.