Forecasting models for monthly energy consumption using machine learning in Burkina Faso

---

Résumé

Lack of forecasting of the electricity consumption in Burkina Faso is of great concern. This leads to selective suspension of power supply due to insufficient production, which has a negative impact on economic activities. In addition, around 68% of production of electricity is supplied by its neigbours such as Ghana and Cote d'Ivoire. The cost of energy is most expensive in the sub-region at 0.21 Euro. We present in this paper, four predictive models of energy consumption to forecast household energy consumption in the Middle West region of Burkina Faso. The proposed models are the Decision Tree Regressor, the Random Forest Regressor, the Neural Network and xgboost. We assessed them through three measures, namely the Root Mean Squared Error, the Mean Absolute Error and the R2- Score. We found that the neural networks have more satisfactory performance. Our contribution is an important step in the planning of energy production to meet the growing consumption of households

Mots-clés

Energy consumption, Biological system modeling, Neural networks, Production, Predictive models, Data models, Planning model, prediction, Machine learning, Energy consumption