Scaling Behavior for Electric Vehicle Chargers and Road Map to Addressing the Infrastructure Gap

Source code and interactive demo for Scaling Behavior for Electric Vehicle Chargers and Road Map to Addressing the Infrastructure Gap.

Getting the Data

The processed dataset, as used in the paper for model fitting, is available at data/dataset.csv. The source datasets, as documented in the supporting information, were collected from the following sources:

Source	Relevant Files	Source
US Bureau of Labor Statistics, Quarterly Census of Labor Statistics	data/blc_qcew	https://www.bls.gov/cew/
Atlas State EV Registration Data	data/ev_registration	https://www.atlasevhub.com/materials/state-ev-registration-data/
US Census Bureau	data/census	https://www.census.gov/
NREL's Alternative Fuel API dataset	data/nrel_ev_stations.csv	https://developer.nrel.gov/docs/transportation/alt-fuel-stations-v1/
State Registration	data/census	Various, See SI

More precise references, particularly for US Census data, are available in the supporting information.

Script for processing the various raw source datasets have also been provided. Please note that this remains a largely manual process. For example, the PA passenger vehicle registration reports were manually transcribed from the source pdf. Further, the data extraction scripts, particularly for state registration data, are highly sensitive to data inputs, and their outputs should be manually verified.

• State Registrations:
  • State-By-State Preprocessing: See the parse* scripts located in the State's folder.
  • Collating States: data/state_registration/extract_passenger_vehicles.py
• EV Registrations
  • Makefile for fetching new results: data/ev_registration/makefile
  • From within data/ev_registration run make all
• BLS Datafiles:
  • Script: data/fetch_bls_data.jl]
  • Makefile for running: data/makefile, run make ./bls_qcew from within data/
• Census Bureau:
  • Shapefiles can be downloaded by running make fetch-data from within data/

Unprocessed Data

Due to size constraints, this repository does not provide the unprocessed data files used to develop our dataset. It can be found at: https://doi.org/10.5281/zenodo.5784659

Replicating the Analysis

Collating the processed datasets (See above), model fitting and plotting was handled by the paper.jl script making heavy use of the code located in src/ using Julia. The code was developed using v1.8.5, and may or may not work with other versions (although we suspect it will).

To run the analysis, from the root directory run the following:

julia --project -L ./paper.jl -e 'using Pkg; Pkg.instantiate(); main()'

Update plots will be saved to img/ and tabulated predictions will be saved to charger_scaling_predictions.csv. Further inspection of the models, fit statistics, and more is possible by entering the Julia REPL as shown below:

> julia --project -L ./paper.jl

julia> fig, models, model_comparisons, df, df_out = main();
[...]
julia> models["county"]["population"] # To inspect the stations vs. population fits
[...]
julia> models["county"]["population"] |> ChargerScale.Models.model_report # Useful summary stats
[...]

Docker Container

We have provided a Docker container for replicating the above analysis. To use, follow the relevant Docker installation instructions for your system. Then:

1. Download and extract the container and complete dataset from: https://doi.org/10.5281/zenodo.5784659
2. Import the container image: docker import evse-scaling-docker.tar.gz evse-scaling:latest
3. Run the container with the following command

docker run --rm -it \
    --volume="$(pwd)/data:/data:ro" \
    --volume="$(pwd)/img-docker:/img" \
    evse-scaling:latest

This will mount the data/ directory as read-only and save any generated images to the img-docker/ directory on the host machine.

Building the Container

The docker image can be built using: docker build -t evse-scaling:latest .