Prediction of Short Term Electric Load Based on BP Neural Networks & ARIMA Combination

Currently, in terms of electric load, single item prediction method is mainly applied based on traditional research methods or artificial neural networks technology. This paper proposes a new method of electric load prediction based on BP neural networks and ARIMA combination model, which first extracts time series data sample in terms of the variation trend of the electric load of Shiqu County, GanZi State, China, and then applies BP neural networks to make correction to the prediction errors, and finally adds the corrected value with ARIMA predicted value. The results show that the mean relative error of the combined prediction method is 0.042, which is of better prediction precision compared to single BP neural networks & ARIMA model prediction method. Thus, to use combination model in the electric load prediction is feasible.