Building Energy Management Systems with LLM Agent Enhanced Natural Language Policy Explanation

Building Energy Management System (BEMS) is important for optimizing energy usage. However, existing BEMS often lack natural language explanations, making it difficult for residents to understand and trust system decisions. This paper proposes a large language model (LLM) Agent-based (BEMS Agent) policy explanation for BEMS, enabling the system to generate comprehensible explanations of environmental parameters and control decisions. To improve reasoning accuracy, we introduce a Reflection-Chain-of-Thought (RCoT) mechanism that enhances the logical consistency of explanations. Furthermore, we develop an automated evaluation method to assess how well users understand the generated explanations. This paper highlights the potential of LLMdriven interpretability in increasing user trust and optimizing energy management efficiency, as evidenced by a 29.4 % improvement in scores when using RCoT prompts compared to CoT prompts.