Application of Elman Neural Network in Top Oil Temperature Prediction of Transformer

Top oil temperature is a key parameter of the transformer in operation, which is difficult to measure and predict. To predict the top oil temperature more correctly, an artificial neural network (ANN) method was proposed. According to field data from many transformers in operation such as ambient temperature, bottom oil temperature, load current and active power, the prediction model based on the Elman neural network algorithm was set up. In the training section, the nonlinear mapping relationships among the top oil temperature, ambient temperature, bottom oil temperature, load current and the active power were determined, which could be employed to directly conduct the prediction of the top oil temperature of transformer with a relatively small error. The confidence interval analysis indicated that the Elman neural network modelling method had high reliability and confidence, and was feasible and accurate enough to predict the top oil temperature of transformer.