Methods of forecasting Alberta's energy demand

Marcelo Guarido, Kristopher A. Innanen

Forecasting energy demand is increasingly important to optimize consumption for residential and commercial tasks. Companies with high demand can control their usage to decrease when energy is most expensive. We used historical data from Alberta’s energy demand and weather data, as energy consumption strongly correlated with temperature. The tested models, linear regression, XGBoost, and Facebook Prophet with XGBoost, presented varying precision when forecasting the energy demand. Linear regression and XGBoost make better predictions on the first week of July 2024, while the Prophet with XGBoost tends to overestimate the demand. Companies should decide which model to use based on the economics of energy price and downtime costs.