Physics-Informed Large Language Models for HVAC Anomaly Detection with Autonomous Rule Generation

2025-10-20

AI Summary

Traditional rule-based methods for HVAC anomaly detection suffer from poor adaptability, while deep learning models lack interpretability and physical consistency; existing LLM-based approaches neglect thermodynamic constraints. Method: This paper proposes a physics-guided LLM-evolution co-design framework. It introduces thermodynamics- and control-theory-informed reflection and crossover operators, enabling the LLM to autonomously generate, evaluate, and iteratively refine interpretable, physically consistent detection rules. A closed-loop optimization system integrates physics-informed modeling, rule-based reasoning, and evolutionary search. Contribution/Results: The framework achieves state-of-the-art detection performance on public benchmarks, producing high-confidence, production-ready diagnostic rules. It significantly enhances transparency, energy efficiency, and reliability of AI-driven building management systems.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Search and Optimization: Learning to Search
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Heating, Ventilation, and Air-Conditioning (HVAC) systems account for a substantial share of global building energy use, making reliable anomaly detection essential for improving efficiency and reducing emissions. Classical rule-based approaches offer explainability but lack adaptability, while deep learning methods provide predictive power at the cost of transparency, efficiency, and physical plausibility. Recent attempts to use Large Language Models (LLMs) for anomaly detection improve interpretability but largely ignore the physical principles that govern HVAC operations. We present PILLM, a Physics-Informed LLM framework that operates within an evolutionary loop to automatically generate, evaluate, and refine anomaly detection rules. Our approach introduces physics-informed reflection and crossover operators that embed thermodynamic and control-theoretic constraints, enabling rules that are both adaptive and physically grounded. Experiments on the public Building Fault Detection dataset show that PILLM achieves state-of-the-art performance while producing diagnostic rules that are interpretable and actionable, advancing trustworthy and deployable AI for smart building systems.

Problem

• Automating HVAC anomaly detection with interpretable, physics-grounded rules
• Overcoming limitations of rigid rule-based and opaque deep learning methods
• Integrating thermodynamic constraints into autonomous rule generation process

Innovation

• Physics-informed LLM framework for HVAC anomaly detection
• Evolutionary loop automatically generates and refines rules
• Physics-informed operators embed thermodynamic constraints