Predicting the error of electronic current transformers is significance for tackling the long-term stability problem of electronic current transformers’ error in time and ensuring the validity of electric power trade. The error model for electronic current transformers is established, where the error of electronic transformers is taken as a theoretical model with one input and multiple outputs, and the input and output variables are determined. Since the relationship between the input and output variables is indistinct, a forecasting method for electronic transformers based on clustering RBF neural network is proposed. The data are pre-processed using Z-score normalization method to avoid the problem of different variable magnitude and unit. The input variables are analyzed by K-means clustering method to simplify the neural network. The numerical example suggests that the predicting error of ratio error is less than 0.05% and the predicting error of phase error is less than 10'. The method provides an effective approach to analyzing the development state of electronic current transformers in operation and to managing the instruments actively, as a result, the risk of electric energy trade can be alleviated.