Research on the prediction-model of the surface-temperature of large oil-immersed power transformer

In order to detect the abnormal temperature of large oil-immersed power transformer earlier, this paper presents a new prediction-model of surface-temperature for the transformer to predict the normal surface-temperature. And to compare the difference between the predicted value and the measured value can judge if the condition of the transformer is fit to operate. The model parameters are estimated from the field data by using the partial-least-square method, which overcomes the instability of results estimated by using the traditional least-square method. Furthermore, the appropriate amount of data for the parameter estimation was decided according to the principle of heat transfer and the characteristics of the predicted deviation, such as mean, variance and normal distribution. The prediction model is verified from the accuracy of results predicted with one year of field data.