Sustainable energy decision-making with an RAG-LLM system

Arslan, Muhammad; Munawar, Saba; Sibilla, Maurizio

Abstract

To reach the ambition of a net-zero economy by 2050, the UK aims to improve energy efficiency in all homes by 2035, targeting reductions in energy consumption and household expenses. Small and Medium-sized Enterprises (SMEs) play a critical part in this evolution but frequently face challenges in navigating the complex and fast-changing Energy sector to access essential information. Required data spanning regulatory updates, market trends, renewable energy production, and climate patterns is dispersed across multiple sources, creating inefficiencies and raising costs for SMEs pursuing sustainable Energy goals. This study addresses this information gap through a prototype Information System (IS) that consolidates diverse regulatory and environmental topics as a proof-of-concept. The proposed Energy Question Answering (QA) Assistant, based on the Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG), integrates information from government, industry, and environmental sources. Using open-source technology, this tool delivers SMEs timely insights into Sustainable Energy Initiatives (SEIs) and regulatory frameworks, supporting cost-effective and informed decision-making that aligns with the UK's sustainability targets.

• Presentation Conference Type: Conference Paper (published)
• Conference Name: 2024 International Conference on Decision Aid Sciences and Applications (DASA)
• Start Date: Dec 11, 2024
• End Date: Dec 12, 2024
• Acceptance Date: Nov 30, 2024
• Online Publication Date: Jan 17, 2025
• Publication Date: Jan 17, 2025
• Deposit Date: Jan 18, 2025
• Publicly Available Date: Jan 21, 2025
• Publisher: Institute of Electrical and Electronics Engineers (IEEE)
• Peer Reviewed: Peer Reviewed
• Pages: 1-5
• Book Title: 2024 International Conference on Decision Aid Sciences and Applications (DASA)
• ISBN: 9798350369113
• DOI: https://doi.org/10.1109/dasa63652.2024.10836639
• Public URL: https://uwe-repository.worktribe.com/output/13625571