Electricity Demand Forecasting Using an Ensemble Model with Feature Selection

Abstract

—Accurate forecasting empowers electricity stake holders to anticipate and arrange operational requirements for electricity production, ensuring reliable power. In this paper, a novel ensemble forecasting model proposed using the outcomes of Random Forest (RF) and Bidirectional Gated Recurrent Unit (Bi-GRU) model are fed to an XGBoost model as input. The input data contains temporal and weather variables,by applying filter, wrapper, and embedded methods a set of common selected fea tures has been identified. With appropriate data pre-processing, it improves the forecasting accuracy. The performance of the proposed model is compared with that of the RF, LightGBM, and XGBoost models, using evaluation metrics as Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Coefficient of determination (R2). The performance metrics of proposed ensemble model is R2 98.84% , RMSE 0.5401, and MAE 0.3738 excels the other comparable models