Various factors that influence power load are more and more intricate. The traditional load forecasting methods can no longer adapt to the situation. Self-organizing method is a comparably new modeling method and so as to be easily used in the recognition and prediction of complex non-linear systems. Compared with traditional forecasting methods, it is of higher prediction accuracy. But common self-organizing methods don't realize the self-organizing function really and cannot show the advantages of self-organizing polynomial. Therefore, combing the artificial neural network with Group Method of Data Handling, a simply-improved self-organizing polynomial algorithm ,which is able to realize the real self-organizing function that is of intelligent characters, is put forward. At last, this kind of self-organizing polynomial algorithm is applied to predict the electric power load in order to prove the validity of this intelligent method as well as that of the improved load forecasting model.