Studies of short-term load forecasting model based on LSTM-NBEATS

Abstract

Due to the current rising in oil prices and energy scarcity, the role of short term load forecasting is critical in basic functioning and scheduling of power systems. The forecasting accuracy of a single model always has its limitations. Therefore, the LSTM-NBEATS model, a combined model combining LSTM and NBEATS by a MAPE (mean absolute percentage error) weighting method is proposed. This model is easy to realize and train, and does not rely on complicated feature engineering. It is applied to hourly load datasets from three European countries, Macedonia (MK), Latvia (LV), and Poland (PL). In this paper, experimental results show that in short term load forecasting the model proposed performs effective.