Abstract: When a single-phase ground fault occurs in the distribution network, the sound operation of the distribution equipment and personal safety will be threatened.Therefore, whether the grounding route can be determined quickly and isolated has become a hot and difficult problem in power system. This paper proposes an online diagnosis model of line grounding based on random forest algorithm. Firstly, the line grounding fault case database is read and pre-processed to achieve data cleaning and feature derivation of electrical characteristics such as voltage, current, active power, reactive power and power factor, then the random forest algorithm is used to calculate the importance of each electrical characteristic and to explore the change law of electrical characteristics of different lines before and after the grounding fault, so as to establish the online diagnosis and analysis model of line grounding. Through cross validation and grid search method, the parameters of the online diagnosis model are optimized, and the intelligent online grounding line selection is realized. Finally, taking the actual distribution network of Shimen Power Supply Company as an example, the online fault diagnosis is carried out by using the trained model, and the results show the effectiveness of the proposed method.