LLMs for EV Infrastructure Permitting

The Electric Vehicle (EV) charging permitting processes' database is a novel, multi-jurisdictional resource designed to contain the required codes and compliances in a structured database. Within this database are three tables, each structured with 287 columns, designed to capture detailed information spanning electrical, structural, zoning, and accessibility aspects, along with data regarding fees, reviews, and process durations. The database contains 99 state-level documents pertaining to 36 U.S. states, in addition to 87 county-level and 101 city-level documents, thus offering a complete overview of guidance and practices regarding permitting. The data was gathered via an Azure-hosted GPT-4o workflow, supplemented by targeted manual Google searches. State and county materials were located and extracted using the GPT-4o model. The Large Language Models (LLM) were used in conjunction with the decision tree framework with targeted prompts to extract the key information. The structured database incorporates Tables 1-3 included below as resources, as well as Table 4 which provides the scores for each document based on the scoring criteria in the paper (to be added after publication). The database can be used to compare and identify the patterns and trends in the requirements across different authorities having jurisdictions. This resource can be used by researchers, policymakers, and project teams. Note: LLMs are known to make mistakes in the interpretation of complex procedural documents and therefore no one should rely solely on this database to inform their own real-world EV infrastructure projects.

[
  "019:20"
]

[
  "019:003"
]