Timesfm and xgboost power day ahead electricity forecasts for university buildings

Machine Learning and Time-Series Foundation Models for Day-Ahead Electricity Forecasting in a University Building

Machine LearningArtificial Intelligence

Summary

Predicting how much electricity a university building will use a day ahead helps manage heating and ventilation efficiently. The authors compare different forecasting methods and find that a pretrained model called TimesFM works best overall, especially when given extra weather information. Another method, XGBoost, also performs well and can be quickly adapted to specific buildings. Including weather details improves predictions for holidays but may slightly affect other days. This work shows that these modern forecasting tools can guide building energy use without heavy local retraining.

What this means in practice

  • For building operators: Use pretrained TimesFM to generate accurate day-ahead electricity demand forecasts to better schedule building heating and ventilation services.
  • For energy management teams: Deploy locally fitted XGBoost models for rapid, adaptable electricity load forecasting in buildings with minimal setup time.

Tested on one dataset.

Authors

Amey Vasulkar, Samuel Kernan Freire, Arie-Willem de Leeuw, Tadeo-Baldiri Salcedo-Rahola

Abstract

Day-ahead electricity forecasts can help building operators schedule heating, ventilation, and other services around anticipated demand. Choosing a forecasting model also requires balancing accuracy with its forecast-time inputs, local training, and computational requirements. We benchmark statistical, machine-learning, recurrent neural-network, and pretrained TimesFM models for a university building. We use chronological validation and evaluate all configurations on the same 57 test days, comparing accuracy across workdays, weekends, and holidays alongside local fitting and forecasting time. Weather covariates use retrospective observations. TimesFM 3.0 with covariates achieves the lowest overall errors, reducing mean absolute error by approximately 43\% relative to weekly seasonal-naive forecasting. XGBoost remains competitive, although paired daily comparisons do not establish a clear mean-absolute-error advantage for TimesFM. TimesFM also performs strongly using only past electricity consumption. Adding covariates improves its holiday forecasts but slightly worsens workday and weekend forecasts, leaving overall mean absolute error almost unchanged. TimesFM requires no building-specific weight training, while XGBoost offers rapid local fitting and inference. The evaluated LSTM takes longer to fit without outperforming either approach. An exploratory historical evaluation reveals seasonal variation, while a separate weather-source comparison produces only small changes in the selected TimesFM configuration's accuracy. These results support pretrained TimesFM as a strong initial forecasting baseline for this building, with XGBoost as a competitive locally fitted alternative. Covariates are particularly useful when improved holiday forecasts justify the additional input preparation.