High-resolution weather-guided surrogate modeling for data-efficient cross-location building energy prediction

A team of researchers has introduced a novel AI approach that could revolutionize how architects and engineers simulate building energy performance. The study, led by Piragash Manmatharasan, Girma Bitsuamlak, and Katarina Grolinger, presents a high-resolution, weather-guided surrogate model designed to predict energy use efficiently across different geographic locations. Traditional methods rely on computationally expensive physics-based simulators like EnergyPlus or require training data from numerous sites to generalize. This new model breaks that constraint by learning short-term, weekly weather patterns that drive energy demand—patterns that recur across regions—allowing it to be trained on data from just one location and still perform accurately elsewhere.

The core innovation is the model's ability to capture and leverage these shared, weather-driven energy signatures. In experiments, when trained on a single site, the model maintained high predictive accuracy for other locations within the same broader climate zone without noticeable performance loss. Even when applied across different climate zones, it exhibited only minimal degradation. This represents a significant leap in data efficiency and scalability over previous weather-informed surrogate models, which needed extensive simulations from multiple sites to achieve similar generalization. The work, detailed in the paper 'High-resolution weather-guided surrogate modeling for data-efficient cross-location building energy prediction,' has been accepted for publication in the journal Energy and Buildings.

This research directly tackles a major bottleneck in sustainable building design: the time and computational cost of running detailed simulations for every potential design in every possible location. By creating a reusable, generalizable surrogate, the method accelerates the design optimization cycle. It enables faster exploration of energy-efficient building forms, materials, and systems for new sites, supporting more rapid adoption of sustainable practices in architecture and urban planning. The model's efficiency makes sophisticated energy modeling more accessible, even for projects with limited data or computational resources.