Abstract: The rapid development of artificial intelligence algorithm and edge computing technology has laid a

foundation for the development of a new type of over-voltage identification device of distribution network with

higher identification rate and stronger intelligence. A design method of hardware and software system for

overvoltage identification in distribution network based on edge computing was proposed. In hardware，a signal

conditioning circuit suitable for high bandwidth and high precision voltage divider was designed by using multistage operational amplifie，the synchronous real-time acquisition of three voltage signals in distribution network was realized by using high-speed data acquisition card，an edge computing data processing unit was constructed based on an embedded processor with neural-network processing unit（NPU） to be used for over voltage data processing and type identification，4G communication was used to transmit overvoltage type and overvoltage fault data. In software，a method of overvoltage identification based on power wavelet transform and convolutional neural network artificial intelligence algorithm was proposed and implemented by using high-performance neural network reasoning engine. High voltage experiment and fault injection method were used to verify the new over-voltage identification device designed for distribution network. The experimental results show that the device can not only realize high-speed and accurate acquisition of over-voltage data，but also the data processing time is shortened and identification accuracy is higher.

      Key words: power distribution network；overvoltage identification；edge computing；convolutional neural

network