Abstract: With the integration of distributed generations and diverse loads, distribution network has changed from passive network to active network, while the power flow has changed from 'unidirectional' to 'bi-directional'. The structure of distribution network, the environment of equipment and the condition of operation are becoming increasingly complex. It is becoming difficult to achieve accurate analysis, identification and location by traditional fault handling methods. Therefore, we discuss the technology challenge of fault diagnosis and location in power distribution network, existing approaches of fault processing, new processing methods based on artificial intelligence. Moreover, we summarize and analyze the domestic and international research status on the issues. Especially, a sample completion method based on time series simulation is proposed to solve the small sample problem. In order to improve the accuracy of training model in fault diagnosis, online fault processing is proposed and discussed through optimal feature extraction, multi-level model training and migration learning optimization. Finally, the problems existing in the current methods and the future research direction are summarized.