EVQUARIUM: An EV charging infrastructure evaluation tool

--- Update 3/1/2026 ---

<img width="1145" height="1020" alt="EVQUARIUM UI Albany" src="https://github.com/user-attachments/assets/dc3a57b2-79bd-409a-9d42-a2864b72c592" /> In this update, we added a user interface "evquarium-osm.hmtl" that can be loaded in Claude to run. It loads up OSM and users can draw a bounding box on the map to define a study area, then load station and census data to model demand and congestion at public charging stations. The tool estimates EV demand by combining vehicle registration counts, an adjustable adoption rate, and a configurable visit frequency, then assigns demand to stations using a Method of Successive Averages (MSA) equilibrium algorithm that minimizes a generalized cost of travel time plus wait time. Station congestion is modeled using M/D/c queuing theory, where each station's service rate is a weighted average of its L1, L2, and DCFC port mix. Results are displayed as a choropleth map of demand origins, flow lines showing assignment volumes, and a ranked table of stations by utilization and expected wait time.

Two screenshots are included, one showing the use of the app for a study area in Northern New Jersey, and one for a study area in Albany.

1 Introduction

EVQUARIUM is an evaluation tool that quantifies the accessibility of EV charging station locations using queueing and graph theory. Given a zonal distribution of EVs with access times to charging stations, it outputs the access patterns and social impacts under equilibrium as expected wait time (queueing+charging) and utilization ratio of each charging station, and average access time (traveling+queueing) of each EV parking location. It could be used to evaluate current or designed EV charging station configurations.

EVQUARIUM does the evaluation by obtaining the assignment matrix at user equilibrium (UE), which describes which charging stations the EVs at each zone choose to go to. At UE, charging stations used by the EVs from the same origin should have the smallest and the same access time (traveling+queueing). Potantial queueing at charging stations is considered through a M/D/C queueing model.

INPUT DATA: zone IDs, charging station IDs (a station with mixed charger types is represented by multiple chargers with same location), number of EV charging visits per unit time per zone, charging time per vehicle from empty to max, travel time between each zone

OUTPUT DATA: queue delay and utilization ratio at each charging station under equilibrium, allocation of EV charging visits from each zone to each charging station under equilibrium

The files include:

"EV Assignment Tool (Example and Instructions) - 20230328.ipynb": the tool and tutorial
"Example input files": input files for the example in the tutorial
"M-D-C approximation": simulation to show the accuracy of the M-D-C approximation we adopt
"NYU Open Source license.pdf": license file

For more details, please refer to the following paper:

Liu, B., Pantelidis, T. P., Tam, S., & Chow, J. Y. (2022). An electric vehicle charging station access equilibrium model with M/D/C queueing. International Journal of Sustainable Transportation, 1-17. https://www.tandfonline.com/doi/full/10.1080/15568318.2022.2029633

For questions about the code, please contact: Bingqing (Chloe) Liu (bingqing.liu@nyu.edu).

2 Instructions

To use this Tool:

Please save the EV location list, charging station list and the travel cost matrix as csv files in the same folder as this script, then fill in the file names in the following cell correspondingly.

Prepare EV location csv file

Columns:

 "ID" : 1,2,3,4,... (int64)

 "Number of EVs" : the number of EVs to be charged per unit time at that location (int64)

Note that in the small example we give in the script, there are extra columns storing the coordinates of the EV locations. These are not necessary for code running.

Prepare Charging Station csv file

Note: Within each charging station, all the chargers are of the same type (Level 2 or DC Fast). If some charging stations have both Level 2 and DC Fast chargers, please seperate each of them as 2 charging stations at the same location.

Columns:
```
 "ID" : 1,2,3,4,... (int64)
 
 "mu": service rate (number of EVs charged/unit time) of one charger at this charging
 station (int64)
 
 "Number of Chargers" : the number of chargers at this charging station (int64)
 
```

Note that in the small example we give in the script, there are extra columns storing the coordinates of the charging stations. These are not necessary for code running.

Prepare travel cost matrix csv file

This matrix should be the travel time cost matrix.

Note:

 1) The sequence of the rows should correspond to the sequence of the EV Locations in the EV location csv file, and the sequence of the columns should correspond to the sequence of the charging station csv file (float64)
 2) No heading and index in the csv file

Set Weights for Access Time and Charging Time

Parameter "access_time_weight": weight of access time. Recommended value is 6.198.

Parameter "charging_time_weight": weight of charging time. Recommended value is 1.

For the recommended values, please refer to Ge 2019.

Reference: Ge, Y. (2019). Discrete Choice Modeling of Plug-in Electric Vehicle Use and Charging Behavior Using Stated Preference Data (Doctoral dissertation).
Set Convergence Criteria

Parameter "e": If the Euclidean norm of the difference between the assignment matrix of this iteration and the last iteration is less than e, the algorithm stops and outputs the results.
Set Printing

Parameter "pri": print error, mean cost and number of steady-state charging stations every pri number of iterations. If no printing is needed before the final results, set pri as inf (pri = np.inf)
Run the algorithm After the algorithm converges with the convergence criterion that you set, please check if all the charging stations are at steady state. The 6 output files are saved in a folder named "Results" where this script is saved.

!!! Please make sure that there's no folder named "Results" at the directiry where the script and input files are saved. The output files include:

 - Assignment Matrix.csv
 - Access Time Matrix.csv
 - Charging Time Matrix.csv
 - Access + Charging Time Matrix.csv
 - Charging Station Table.csv
     - Columns:“Utilization Ratio”, 
               “Expected Queue Delay”, 
               “Charging Time”, 
               “Expected Queue Delay + Charging Time”
 - EV Parking Location Table.csv
     - Columns:“Average Access Time”, 
               “Average Charging Time”, 
               “Average Access +Charging Time”

The following results are directly printed:
 - System Total Access Time
 - System Total Access Time + Charging Time
 - Average Access Time for one EV
 - Average Access Time + Charging Time for one EV
 - Number of steady-state (utilization ratio ≤ 1) charging stations at convergence 
 - Number of iterations until convergence
 - Run time of the algorithm (wall time)

!!! Please check if the number of steady-state charging stations at convergence equals to the total number of charging stations. If a large number of charging stations are not at steady state, the results are not accurate. Run the algorithm below:

3 Example

We use a small example with 5 EV parking locations (demand nodes) and 3 charging stations to illustrate how the example works.

EV parking location table:

Charging station table:

Their locations plotted:

The travel cost matrix is computer using linear distance assuming a speed of 20.

Convergence criterion is set as 10^(-4). Printing setting is to print every 10000 iterarions.

Output:

The output includes error, mean cost of accessing and charging, and number of charging stations at steady state every 10000 iterations. Final assignment matrix and travel+queueing+charging table are printed afterwards.

Convergence is reached after 36869 iterations. Computation time is 11.1 sec.

From the results, we can see Wardrop's principles. For a parking location, the chosen charging stations have the smallest and the same cost of travel+queueing+charging.

Detailed results can be illustrated as follows. For parking locations, average cost of accessing assigned charging stations is computed.

For charging stations, utilization ratio and average queueing+charging time is computed.

<img width="371" alt="Screen

EVQUARIUM

Install / Use

README

EVQUARIUM: An EV charging infrastructure evaluation tool

1 Introduction

2 Instructions

3 Example