Optimizing EV Charger Locations in Northwest Arkansas

AMPL modeling
Author

Fernando Nunez, Drake Wolfenden

Published

December 7, 2025

Introduction to Project

This project aims to optimize the placement of electric vehicle charging stations across Northwest Arkansas by identifying the best set of station locations that minimize travel distance for residents. Our goal was to recommend the most efficient charging network for the region.

Group Members

  • Fernando Nunez
  • Drake Wolfenden

Skills Learned

  • AMPL modeling & debugging
  • Model Formulations

Project Development Process

Our original goal was simply to find the best locations for EV chargers in Northwest Arkansas. The scope evolved as we realized we could provide more insight by running multiple values of p instead of only one. Along with that, some major roadblocks included aligning AMPL indexing, cleaning the dataset, and debugging incorrect code. The final model offered good recommendations about the optimal number and placements of EV stations in Northwest Arkansas.


Key Features or Highlights

1. Methodology

Decision Variables

  • yⱼ — whether a charging station is built at ZIP code j
  • xᵢⱼ — proportion of demand from ZIP i assigned to station j

Parameters

  • wᵢ — population weight for ZIP i
  • dᵢⱼ — distance between ZIP i and ZIP j
  • p — number of stations allowed (3, 5, or 7)

Objective Function Minimize total population-weighted travel distance: \[ \ \text{Minimize } \sum_{i=1}^{n} \sum_{j=1}^{n} w_{i} \, d_{ij} \, x_{ij} \ \]

Core Constraints

  • Every ZIP must be assigned entirely.
  • ZIPs can only assign demand to locations where yⱼ = 1.
  • Exactly p stations must open.

2. Data Collection & Parameter Construction

Data Source: SimpleMaps (2024) U.S. ZIP Code Database
https://simplemaps.com/data/us-zips

Steps Followed

  • Filtered ZIP codes to Benton & Washington Counties
  • Used coordinates to build distance matrix
  • Extracted ZIP code populations to determine demand weights
  • Processed data into AMPL

Results & Scenario Analysis

We tested three expansion levels: p = 3, p = 5, and p = 7.

Baseline Scenario — p = 5

  • Total Population: 541,371
  • Average Distance: 4.56 miles

Selected ZIPs:

  • 72701
  • 72704
  • 72712
  • 72758
  • 72764

These correspond to Fayetteville, Springdale, Rogers, and Bentonville


Scenario: p = 3

  • Average Distance: 6.87 miles
  • Major areas must travel farther
  • Coverage is weaker, especially for rural ZIPs

Scenario: p = 7

  • Average Distance: 3.23 miles
  • Additional stations reduce travel inequality
  • Best for rural ZIP accessibility, but diminishing returns appear

Diminishing Returns Table

p Avg Distance
3 6.87 miles
5 4.56 miles
7 3.23 miles

The biggest improvement occurs from p = 3 → p = 5. Increasing beyond 5 stations still helps rural areas but yields smaller overall gains.


Reflection

This project helped me understand how an optimization model works in a real situation. At first, AMPL and the data setup were confusing, but after working through the steps it started to make more sense. I learned how important it is to clean the data and correct indexing are, because small mistakes can break the whole model. In helping build the distance matrix and testing the scenarios, it showed me how changing one parameter can affect results drastically. Overall, this project helped me get better at problem-solving, and feeling more confident in reading solver output.