Large Language Model-Assisted Planning of Electric Vehicle Charging Infrastructure with Real-World Case Study

TL;DR

The paper tackles city-scale EV charging infrastructure planning by jointly optimizing long-term investment and short-term EV assignment under spatial-temporal demand. It introduces an LLM-assisted modeling workflow to rapidly generate and refine the optimization formulation and pairs it with a distributed ADMM solver to achieve scalable, city-scale computation. A Chengdu case study with 1.5 million real travel records demonstrates a 30% reduction in total cost compared with a baseline that ignores EV assignments, and the ADMM approach yields roughly two orders of magnitude faster computation than centralized solving. The work advances practical, scalable planning by integrating human-in-the-loop LLM modeling with decomposed optimization, enabling more cost-effective and efficient EV charging networks on standard hardware.

Abstract

The growing demand for electric vehicle (EV) charging infrastructure presents significant planning challenges, requiring efficient strategies for investment and operation to deliver cost-effective charging services. However, the potential benefits of EV charging assignment, particularly in response to varying spatial-temporal patterns of charging demand, remain under-explored in infrastructure planning. This paper proposes an integrated approach that jointly optimizes investment decisions and charging assignments while accounting for spatial-temporal demand dynamics and their interdependencies. To support efficient model development, we leverage a large language model (LLM) to assist in generating and refining the mathematical formulation from structured natural-language descriptions, significantly reducing the modeling burden. The resulting optimization model enables optimal joint decision-making for investment and operation. Additionally, we propose a distributed optimization algorithm based on the Alternating Direction Method of Multipliers (ADMM) to address computational complexity in high-dimensional scenarios, which can be executed on standard computing platforms. We validate our approach through a case study using 1.5 million real-world travel records from Chengdu, China, demonstrating a 30% reduction in total cost compared to a baseline without EV assignment.

Figures (7)

• Figure 1: An illustrative example of an EV charging infrastructure planning scenario, showing varying traffic flows, charging demand, and planning costs across urban locations, and a coupled decision framework linking investment and assignment.
• Figure 2: Contrasting traditional and LLM-assisted approaches to the modeling process.
• Figure 3: An illustrative example of the computing and communication process for distributed ADMM.
• Figure 4: Primal and dual errors over iterations for the ADMM approach.
• Figure 5: Optimal investment decisions in maps showing the distribution of the investment decisions. The bubbles correspond to the locations where the area of the bubbles represents the amounts of charging capacities. The area of bubbles of the two models has been separately scaled according to the proportion of total investment decisions. The color represents the location-dependent investment cost for different locations.