Short-term electricity consumption forecasting based on the attentive encoder-decoder model

Electricity consumption forecasting plays an important role in establishing and maintaining electric supply management systems. Power companies need to keep a balance between the power demand and supply for customers; this requires an accurate forecast. However, electricity consumption forecasting is affected by various factors such as different weather conditions, season, or temperature. If we cannot predict electricity accurately, the balance between the demand and supply would be destroyed, which may cause huge penalties to power companies. Therefore, electricity consumption forecasting is an important task. The purpose of this study was to forecast the electricity consumption of a manufacturing company every half an hour in the next day to prevent a power supply company from running out of power. In our work, we proposed a short-term electricity consumption forecasting method based on the attentive encoder-decoder and several nonlinear multi-layer correctors. The proposed method is verified in several experiments by using the actual data on electricity consumption of the manufacturing company. The results show that the proposed method outperforms previous methods.