Electricity Sales Forecasting Based on Model Fusion and Prophet Model.

Accurate forecast of electricity sales is a very meaningful task for both electricity companies, security and government departments. This paper proposes a forecasting model for short-term, mid-term and long-term electricity sales respectively. For short-term and mid-term forecasting, we use multiple base models to make predictions and use model fusion methods to get the final prediction results. As for long-term forecasting, the tuned Prophet is used to make a prediction. Through experiments, we found that XGBoost as the base model can achieve the best prediction effect, in which the short-term prediction error can reach 1.97% and the average error of the mid-term prediction is 1.472%. In the long-term forecasting, the MAPE of the entire month of June 2020 is 3.64%. It can be seen that they have achieved good prediction results.