Abstract: Aiming at the problems of frequent failures and untimely maintenance of DC charging pile，a fault

early warning method of DC charging pile based on soft classification model was proposed. Based on the current

sequence and voltage sequence of the charging pile in the charging process，the differential sequence of charging

power was constructed，the state of charge（SOC）sequence of the electric vehicle was also regarded as the feature of the charging process，above characteristics were taken together with the basic information of the charging pile as the input feature of the fault identification model；based on the integration of random forest（RF），extreme gradient boosting（XGBoost）and K-nearest neighbor（KNN）algorithms，a charging pile fault identification model was constructed，the soft classification method was used in the model to calculate the probability that the charging characteristics of  the charging pile belong to various fault states. By setting the early warning threshold，when the probability of failure of the charging pile exceeded the early warning threshold，an early warning message was sent to the relevant maintenance personnel. The DC charging pile data obtained from the internet of vehicles platform was used for experiments，the results show that the proposed method can carry out early warning for the DC charging pile that is about to fail，and the precision and recall of early warning meet the requirements，which verifies the effectiveness and practicability of this method. The proposed method has certain practical value for the operation and maintenance of DC charging pile.

    Key words: DC charging pile；fault early warning；soft classification；ensemble model