DemandCast: Global hourly electricity demand forecasting

TL;DR

DemandCast tackles the challenge of forecasting global hourly electricity demand under diverse regional data conditions. It combines historical demand with weather and socioeconomic predictors using XGBoost to produce the normalized demand profile $D_n(t)$ across 56 countries (2000–2025). The study delivers a scalable, open-source end-to-end pipeline with a strict temporal-splitting evaluation, achieving a test-set MAPE of $9.2\%$ and highlighting data-availability disparities. The work provides a practical tool for energy planners and policymakers, supporting scenario analysis and grid management amid the global energy transition.

Abstract

This paper presents a machine learning framework for electricity demand forecasting across diverse geographical regions using the gradient boosting algorithm XGBoost. The model integrates historical electricity demand and comprehensive weather and socioeconomic variables to predict normalized electricity demand profiles. To enable robust training and evaluation, we developed a large-scale dataset spanning multiple years and countries, applying a temporal data-splitting strategy that ensures benchmarking of out-of-sample performance. Our approach delivers accurate and scalable demand forecasts, providing valuable insights for energy system planners and policymakers as they navigate the challenges of the global energy transition.

Figures (4)

• Figure 1: Countries and subdivisions for which retrieval modules of electricity demand data are available.
• Figure 2: Training, test, and validation sets in which available electricity demand was split.
• Figure 3: Comparison between historical and forecast electricity demand.
• Figure 4: MAPE values resulting from the testing of DemandCast.