Data-Driven Optimization of EV Charging Station Placement Using Causal Discovery

TL;DR

The paper tackles the problem of optimally placing electric vehicle charging stations by embedding empirical charging behavior into a causal discovery and optimization pipeline. It analyzes charging events from Palo Alto and Boulder to uncover that utilization is governed primarily by three factors: proximity to amenities, EV registration density, and adjacency to high-traffic routes, a finding that remains consistent across algorithms. A two-stage framework combines DAG-based structural learning (via NOTEARS and DAGMA) to reveal dependency structures, followed by a Bayesian regression and binary optimization to identify high-utility candidate locations. The approach yields actionable, cluster-oriented site recommendations and demonstrates how data-driven causal insights can enhance station utilization and user convenience, with implications that adapt to different stages of EV adoption. This framework offers a scalable, evidence-based pathway for cities to tailor charging infrastructure to real usage patterns rather than relying on uniform distribution models.

Abstract

This paper addresses the critical challenge of optimizing electric vehicle charging station placement through a novel data-driven methodology employing causal discovery techniques. While traditional approaches prioritize economic factors or power grid constraints, they often neglect empirical charging patterns that ultimately determine station utilization. We analyze extensive charging data from Palo Alto and Boulder (337,344 events across 100 stations) to uncover latent relationships between station characteristics and utilization. Applying structural learning algorithms (NOTEARS and DAGMA) to this data reveals that charging demand is primarily determined by three factors: proximity to amenities, EV registration density, and adjacency to high-traffic routes. These findings, consistent across multiple algorithms and urban contexts, challenge conventional infrastructure distribution strategies. We develop an optimization framework that translates these insights into actionable placement recommendations, identifying locations likely to experience high utilization based on the discovered dependency structures. The resulting site selection model prioritizes strategic clustering in high-amenity areas with substantial EV populations rather than uniform spatial distribution. Our approach contributes a framework that integrates empirical charging behavior into infrastructure planning, potentially enhancing both station utilization and user convenience. By focusing on data-driven insights instead of theoretical distribution models, we provide a more effective strategy for expanding charging networks that can adjust to various stages of EV market development.

Figures (10)

• Figure 1: CS locations with charging consumption level in Palo Alto and Boulder.
• Figure 2: Visualizations for traffic flow proximity (left) and network centrality (right) exemplary for Palo Alto. In the left map the highlighted streets mark traffic arteries. CS locations within 70 meters straight line distance of a traffic artery are marked green, and CS locations further than that are marked red. In the right plot green (red) stands for a high (low) network centrality.
• Figure 3: Comparison of performance of NOTEARS and DAGMA on raw vs. scaled data. For the case of the scaled data $W_{est}$ is interpreted as defining a DAG as well as defining an undirected graph.
• Figure 4: SHD vs. the number of reversed edges on scaled data(R for reversed).
• Figure 5: consumption level by amenities vs. shopping and food (left) and consumption level by high traffic vs. ev count (right)