Error prediction method of electronic transformers based on RBF neural network

With the gradual application of electronic transformers in the intelligent substation, the influence of the typical disturbing source on the electronic transformers is gradually to the industry's attention. In order to ensure the accuracy and reliability of electronic transformers, it is necessary to determine the error of the electronic transformers accurately through the study of the influence factors of the measurement error of the electronic transformers. In this regard, this paper proposes a transformer error prediction method which combines RBF neural network prediction with K-Means clustering algorithm. First, the historical data are classified by K-Means clustering. Then, train the prediction model with the historical data clustered. Take the influencing factors such as temperature, load capacity and magnetic field as the independent variables and the error as the prediction target to establish the prediction model. The prediction model is suitable for the analysis of a large number of data and can accurately predict the error. After using this model to analyze the long-term experimental data of Dongshan substation in 2016, the relative error of the error result is around 5%, which is relatively and verifies the validity of the model.