Supporting energy policy research with large language models: A case study in wind energy siting ordinances

The recent growth in renewable energy development in the United States has been accompanied by a simultaneous surge in renewable energy siting ordinances. These zoning laws play a critical role in dictating the placement of wind and solar resources that are critical for achieving low-carbon energy futures. In this context, efficient access to and management of siting ordinance data becomes imperative. The National Renewable Energy Laboratory (NREL) recently introduced a public wind and solar siting database to fill this need. This paper presents a method for harnessing Large Language Models (LLMs) to automate the extraction of these siting ordinances from legal documents, enabling this database to maintain accurate up-to-date information in the rapidly changing energy policy landscape. A novel contribution of this research is the integration of a decision tree framework with LLMs. Our results show that this approach is 85 to 90 % accurate with outputs that can be used directly in downstream quantitative modeling. We discuss opportunities to use this work to support similar largescale policy research in the energy sector. By unlocking new efficiencies in the extraction and analysis of legal documents using LLMs, this study enables a path forward for automated large-scale energy policy research.