Evolution of LPG Demand Using Machine Learning Planning Models: An Application in the Case of Morocco

The increase in energy demand coupled with threats of climate change, underlines the need to provide a rigorous forecasting energy model in order to manage energy policies effectively. In this paper, we employed three popular machine learning tools to accurately project liquefied petroleum gas (LPG) consumption in the Moroccan residential sector, using data spanning from 1990 to 2017. These tools include multiple linear regression (MLR), support vector regression (SVR) and an artificial neural network approach (ANN). Comparison between models based on MLR, SVR and ANN was done for the period 2018–2020 using statistical parameters. Results show that the ANN model is more accurate and generalized for prediction of LPG consumption in Moroccan residential sector. In effect, Morocco will consume around 3928 Ktoe of LPG in 2030. More efforts should be devoted to improving residential energy efficiency through technological progress, and investments need to be made in renewable energies to reduce overdependence of residential sector on fossil fuels.