Simplified model for the short-term forecasting of heat loads in buildings

Abstract

A data-driven model is used to predict one-hour ahead heat loads based on present and recent history of weather and heat loads. A computationally inexpensive method is built to deliver load forecasting based on existing data quality and resolution from smart meters. Optimal model formulation is discussed and optimized at 4-hour historical values. The model is trained and tested against synthetic data from a building energy simulation, resulting in absolute error <4% and R2 values in the range of 0.92 to 0.94.

Keywords

• Regression model
• Heating demand
• Building
• Lagged value