ADAPT: Physics-Aware Diffusion-based World Models for Adaptive Predictive Transferable HVAC Control

arXiv:2608.198042026-08-21

건물이 계절과 지역을 바꿔도 안정적으로 작동하는 냉난방 AI, 물리 법칙을 함께 학습시켜 만들었다

건물 냉난방(HVAC)은 전 세계 에너지 소비와 이산화탄소 배출의 약 3분의 1을 차지하는데, 벽이나 바닥이 열을 저장했다가 늦게 내놓는 '열 관성' 때문에 제어가 어렵고 계절이나 지역이 바뀌면 기존 AI 제어기의 성능이 크게 떨어진다. 연구진은 확산 모델(diffusion model)이라는 생성 AI에 건물의 열평형 방정식을 결합한 ADAPT라는 예측 모델을 만들어, 실내 온도가 앞으로 어떻게 변할지 미리 예측하고 이를 강화학습 제어기에 알려주도록 했다. 그 결과 같은 조건에서는 기존 최고 성능 대비 에너지 소비 7.3%, 재실자 불쾌감 30.2%를 줄였고, 학습에 없던 계절이나 기후 지역으로 옮겨도 성능 저하가 거의 없었다.

논문 원문 초록 (영문)

Buildings account for roughly one-third of global energy consumption and CO$_2$ emissions. Optimizing indoor climate systems plays a critical role for urban climate mitigation aligned with UN Sustainable Development Goals 11 and 13. However, indoor delayed thermodynamic responses and partial observability severely hinder existing methods, which are primarily limited by implicit thermal inertia, occupancy dynamic prediction, and cumulative prediction errors, especially for out-of-distribution environments. In practice, these challenges are further exacerbated by the high cost and privacy burden of dense indoor sensing, forcing operators to collect only limited data in a single operating regime while expecting controllers to generalize reliably across unseen seasons and climate regions. To address this problem, we propose ADAPT, a physics-aware conditional diffusion indoor environmental world model for HVAC control. The model predicts a short-horizon held-action thermal baseline to capture the latent thermal inertia of the buildings. The diffusion backbone utilizes the robustness of generative models, while a learnable multi-zone heat-balance regularizer constrains generated trajectories to satisfy transferable building thermodynamics without requiring known building geometry or manually calibrated thermal parameters. A credit assignment is then design for the downstream reinforcement learning. Extensive experiments on SemibuildingSim and Sinergym demonstrate that ADAPT reduces HVAC energy consumption by 7.3\% and occupant discomfort by 30.2\% compared with state-of-the-art baselines under IID control. Under OOD control scenarios spanning unseen seasons and climate regions, ADAPT maintains robust performance with only marginal degradation relative to its IID performance, substantially outperforming existing methods in transfer robustness.

저자 · Xu Yang, Kailai Sun, Dianyu Zhong, Qianchuan Zhao

그림 출처: Xu Yang et al., arXiv:2608.19804, CC BY 4.0