Balancing cost and carbon in Manchester non-domestic building retrofits with multi-objective reinforcement learning and large language model

View at publisher Summarise with VIRAT
Authors: Ziru Tao, Paul Baguley, Rashid Maqbool, Obuks Augustine EJOHWOMU
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

The retrofit of non-domestic buildings in the United Kingdom (UK) requires balancing upfront investment against long-term carbon reduction targets, a complex task exacerbated by the need to incorporate the social cost of carbon (SCC) within public-sector appraisal rules. This study addresses the lack of standardised methodologies for integrating Artificial Intelligence (AI) into cost-engineering workflows under such constraints. It develops and tests a hybrid AI decision-support framework that combines Multi-Objective Reinforcement Learning (MORL) with a constrained Large Language Model (LLM) for city-scale retrofit planning. Using Manchester's non-domestic Energy Performance Certificate (EPC) stock, the framework formulates retrofit sequencing as a sequential decision problem, embedding Green Book discounting and Royal Institution of Chartered Surveyors (RICS) Whole Life Carbon Assessment (WLCA) boundaries. The MORL agent generates cost-carbon Pareto frontiers for representative office archetypes, revealing non-linear trade-offs and ‘knee’ points where modest cost increases yield substantial emission savings. The LLM assists in extracting structured parameters from policy documents and translates MORL outputs into stakeholder-facing narratives. Results demonstrate that the framework identifies SCC-optimal retrofit pathways, outperforms classical baselines in exploring the solution space, and provides a transparent, auditable process for embedding AI in regulated appraisal contexts. This study offers a replicable model for using hybrid AI to extend cost-engineering capabilities, supporting more informed investment decisions that align financial and societal objectives.

Fields of Study & Subjects

Building Energy and Comfort Optimization Building energy efficiency and sustainability Life Cycle Costing Analysis

Publication details

Publication date
23 March 2026
Language
EN
Document type
Preprint