Poster Abstract: RE-STEP: A GenAI Prospecting Tool to Assess Community-Based Factors in Early-Stage Renewable-Energy Siting

Charles Alba

Abstract: Renewable-energy developers often rely on prospecting tools to consider a variety of factors during site prospecting. Community-based factors - such as zoning, ordinances, and community sentiment - are challenging to evaluate because they change frequently. Up-to-date local-government documents can indicate a community's regulatory posture and potential receptivity to new renewable-energy projects. When coupled with generative AI - specifically retrieval-augmented generation (RAG) - these sources can be analyzed at scale by utilities and developers. However, rhetorical nuances in these documents—such as legalistic register and euphemistic framing that mask underlying intentions (e.g., opposition to wind-turbine projects framed as wildlife protection, citing avian mortality) - make traditional RAG approaches brittle. Hence, this work proposes Reason-based RAG for Early-Stage Developers for Community-based Prospecting (RE-STEP), a novel RAG tool that uses reason-based retrieval and ExpertPrompting-style generation to retrieve developer-relevant text and generate developer-centric recommendations, respectively. Demonstrations on six local-government documents show that RE-STEP outperforms traditional RAG baselines, particularly on euphemistically framed texts that mask underlying intentions and bury action-bearing signals. This proof-of-concept demonstrates RE-STEP's utility as a prospecting tool for assessing community-based factors, helping utilities mitigate wasted early-stage investment on sites likely to face delays or cancellations due to community or legal pushback. The next logical step is to compile a large database to evaluate RE-STEP at scale.

External IDs:doi:10.1145/3736425.3772105