Closing the Expertise Gap in Residential Building Energy Retrofits: A Domain-Specific LLM for Informed Decision-Making

2026-02-19

AI Summary

This study addresses the challenge faced by residential homeowners in formulating effective energy retrofit strategies due to limited technical expertise. The authors propose the first domain-specific approach that integrates physics-based building simulation with a large language model (LLM) to automatically generate high-accuracy retrofit recommendations and performance predictions from basic user-provided housing information. By efficiently fine-tuning the LLM using LoRA and leveraging simulation and techno-economic data from 536,416 U.S. residential archetypes, the method successfully maps non-expert descriptions to precise technical interventions. Experimental results demonstrate that the top-three recommendations include the optimal carbon-reduction solution in 98.9% of cases and the shortest payback-period option in 93.3%, while reducing CO₂ prediction error by an order of magnitude and significantly improving the accuracy of energy use and cost forecasts.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Search and Optimization: Sampling/Simulation-based Search
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Large language models for search
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Residential energy retrofit decision-making is constrained by an expertise gap, as homeowners lack the technical literacy required for energy assessments. To address this challenge, this study develops a domain-specific large language model (LLM) that provides optimal retrofit recommendations using homeowner-accessible descriptions of basic dwelling characteristics. The model is fine-tuned on physics-based energy simulations and techno-economic calculations derived from 536,416 U.S. residential building prototypes across nine major retrofit categories. Using Low-Rank Adaptation (LoRA), the LLM maps dwelling characteristics to optimal retrofit selections and associated performance outcomes. Evaluation against physics-grounded baselines shows that the model identifies the optimal retrofit for CO2 reduction within its top three recommendations in 98.9% of cases and the shortest discounted payback period in 93.3% of cases. Fine-tuning yields an order-of-magnitude reduction in CO2 prediction error and multi-fold reductions for energy use and retrofit cost. The model maintains performance under incomplete input conditions, supporting informed residential decarbonization decisions.

Problem

• energy retrofit
• expertise gap
• residential buildings
• informed decision-making
• technical literacy

Innovation

• domain-specific LLM
• energy retrofit
• Low-Rank Adaptation (LoRA)
• physics-based simulation
• residential decarbonization