AI Driven Transformation of Building Assessment Report into Energy Models for Building Portfolio Analysis
Bibliographic record
Abstract
Building energy modeling is essential for achieving energy optimization and net-zero targets, yet portfolio-scale analysis faces significant constraints due to critical building information remaining trapped within unstructured documentation such as permits, energy audits, and maintenance records that cannot be systematically extracted and analyzed. This research presents an AI-driven framework utilizing local Large Language Models (LLMs) with Retrieval-Augmented Generation (RAG) to automatically extract and structure building information while ensuring complete data privacy and security. The methodology employs on-premises LLM deployment to process sensitive documentation without external data transmission. This approach efficiently transforms diverse unstructured building data into standardized inputs for the Honeybee Python modeling platform, automatically generating detailed energy models and producing comprehensive portfolio-wide performance analytics. The resulting portfolio-wide analytics identify key patterns, inefficiencies, and optimization opportunities across building inventories, while the detailed energy models provide enhanced baseline representations that can be further refined by energy modelers or improved through more sophisticated RAG implementations for deeper analysis. Validation across 80 Quebec office buildings demonstrated 92.5% processing reliability while reducing analysis timeframes from days or weeks to just a few hours—representing substantial improvements in portfolio assessment efficiency. The framework successfully addresses persistent data processing limitations that have constrained evidence-based energy management, enabling the systematic, data-driven portfolio strategies necessary for achieving ambitious decarbonization objectives.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.009 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.010 | 0.005 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".