The development of electronic transformers is becoming faster with the development of intelligent substation technology. Although it has better improvement in stability, safety and other performance than traditional transformers, there are still some difficulties in the aspects of measurement situation monitoring, status and error evaluation. In order for electronic transformers to be better put into engineering use, their operational reliability must be well evaluated. In this paper, Xgboost is used to model the operation error of the electronic transformer, and the result of the model is processed to obtain the operational reliability probability of the electronic transformer. First, the collected data is clustered by the DBSCAN density clustering algorithm. Then the reliability probability is calculated for each cluster, and then the reliability probability is used as a label, and the collected data is used as an input to establish an Xgboost model. The trained model can make a probabilistic assessment of the operational reliability of an electronic transformer without the measurement results of traditional electronic transformers as standard.