Stacking Model for Short-Term Electrical Load Forecasting

The successful implementation of electrical load forecasting is essential for achieving emission reduction goals and ensuring an optimal equilibrium between power generation and consumption. This paper introduces a load forecasting technique that employs the stacking ensemble method for combining multiple models. Most appropriate basic models were selected in order to maximize the performance of the stacking model. The algorithm used to determine the selection of the model in the second layer is designed to mitigate the risk of overfitting. A selection of load data that accurately represents the system's behavior is chosen to assess the algorithm's viability. The results suggest that the stacking learning framework improves the overall prediction accuracy by reducing parameters such as mean absolute percentage error (MAPE), root mean square error (RMSE), and mean absolute error (MAE).