Long-Term Modeling and Forecasting of the Load at a Substation in Lomé Zone A, Togo

Abstract

Developing an econometric model for estimating long-term electricity demand in Lomé Zone A until 2023 is essential for effective energy planning, enabling optimized demand management and efficient production strategies. Given the complexity of factors affecting electric load, we introduce a hybrid ARIMA/GARCH model, combining ARIMA, GARCH, and LSTM for more precise predictions. This approach facilitates the selection of the optimal model based on performance and reliability. Results show that both ARIMA/GARCH and LSTM models yield good validation results, with the ARIMA/GARCH model excelling in generating realistic predictions, boasting root mean squared errors and mean absolute errors of 17.798 and 16.459, respectively. This highlights the reliability of the ARIMA/GARCH model, providing a solid foundation for enhanced energy management and planning in this growing region.