A Multi‑agent Rag‑enabled Knowledge System for National‑scale Whole‑life Carbon Assessment of Residential Retrofits

Abstract: National residential renovation is critical to meeting climate targets, making Whole-Life Carbon Assessment (WLCA) essential for balancing operational and embodied emissions. National-scale WLCA remains challenging due to fragmented datasets, incomplete material inventories, and reliance on expert interpretation, limiting scalability and decision-making efficiency. This study proposes a multi-agent Retrieval-Augmented Generation (RAG)- enabled Knowledge-Based System (KBS) for national-scale WLCA of residential retrofit scenarios. The framework integrates national data acquisition, building archetype modelling, parametric LCA simulation, a RAG-supported knowledge base, and a stakeholder dashboard. Domain-specific LLM agents covering operational, embodied, and whole-life carbon are coordinated by an interface agent to retrieve, validate, and explain carbon metrics with full data provenance. This improves transparency and interpretability while reducing effort. Case studies on Irish semi-detached dwellings demonstrate strong performance, achieving 88% feasibility, 81% nDCG, 70% precision, and 79% mean reciprocal rank, highlighting the potential of multi-agent LLM systems for evidence-based national decarbonisation planning.