Optimization of Short-Term Forecast of Electric Power Demand in the city of Yaoundé-Cameroon by a hybrid model based on the combination of neural networks and econometric methods from a designed energy optimization algorithm

Abstract

Electrical load forecasting has become a very important task for the management and planning of electrical energy. Several methods have been developed in the literature to solve this task, but researchers continue to search for a stable model that minimizes the prediction error as much as possible. In this article, we propose a hybrid model of artificial intelligence and statistical methods, designed from an optimization algorithm for hourly forecasts of electricity over a period of one week. To do this, we started with a comparative study of predictions with models of artificial neural networks (ANN), multiple linear regression (LRM) and Holt exponential smoothing (HES). Then we retained the best parameters of each model to design our hybrid model (ANN-LRM-HES). The results obtained by each model are considered acceptable in view of the statistical indicators. These results show that apart from Deep Learning techniques which offer excellent results the hybrid model (ANN-LRM-HES) provides the best prediction results, followed by the ANN, LRM and HES models respectively. From the comparative study it emerges that the proposed hybrid model outperformed most similar models in the literature on the subject, obtaining statistically significant precision values.