Toward a foundational thermal model for residential buildings

arXiv cs.LG / 5/5/2026

📰 NewsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper argues that the building energy community needs a single “foundational” thermal model that can generalize across different residential buildings, climates, and control strategies without per-building calibration.
It proposes a physics-informed, decoder-only transformer that injects thermal-domain knowledge via derivative enrichment and Euler-based numerical integration, while using static building features and rotary position embeddings for temporal modeling.
On the CityLearn dataset (247 residential buildings across three climate zones), the model reaches strong one-step prediction accuracy with RMSE around 0.29–0.30°C and improves over both traditional baselines and fine-tuned time-series foundation models.
The model shows zero-shot transfer: training on as few as two buildings can generalize to unseen buildings and climate zones without additional fine-tuning, suggesting a path toward universal building thermal modeling.
The authors note that results are limited to simulated residential buildings, but conclude that physics-informed architectural principles are a promising basis for future universal thermal foundation models.

Abstract

The building energy community lacks a foundational thermal model, i.e., a single pretrained model capable of generalizing across diverse buildings, climates, and control strategies without building-specific calibration. Achieving this vision requires architectural principles that capture universal thermal dynamics rather than memorizing building-specific patterns. We take a step toward this goal by presenting a physics-informed transformer architecture that embeds domain knowledge, e.g., derivative enrichment and Euler-based numerical integration, into a decoder-only framework. We incorporate static building features extracted from simulation models and employ Rotary Position Embedding attention to capture temporal dependencies. Evaluated on the CityLearn dataset spanning 247 residential buildings across three climate zones, our model achieves one-step prediction accuracy (RMSE of 0.30{\deg}C in Texas, 0.29{\deg}C in Vermont) while outperforming both traditional baselines and fine-tuned Time-Series Foundation Models. We also demonstrate zero-shot transferability: models trained on as few as two buildings generalize to unseen buildings and climate zones without fine-tuning. Despite the limitation of simulated residential buildings, our results establish physics-informed architectural principles as a promising foundation for universal building thermal models.