Introduction
With the growing emphasis on clean transportation, the demand for electric vehicles (EVs)
is increasing. However, there currently are not
enough EV charging stations to meet this
demand. Many of the existing chargers are concentrated in specific areas, leaving large
parts of
the city underserved. This imbalance can lead to equity issues, range anxiety,
and congestion around existing chargers.
EVCS-OPTIM seeks to address these challenges while also optimizing placement to keep
energy costs affordable for ratepayers. Our project
uses data-driven methodologies,
including geospatial analysis and optimization modeling, to find the most effective
parking lot locations for
new EV charging stations within San Diego County. By
integrating previous EV adoption data, traffic patterns, and proximity information,
this
project provides an interactive tool to assist San Diego Gas & Electric (SDG&E),
which owns and operates thousands of EV charging stations
across the region, in
strategically expanding its EV infrastructure.
Methodology
Data Analysis
We conducted correlation analysis to determine what factors influence the location of current EV charging locations. Based on this analysis, we selected the following features: distance to the nearest charger, number of chargers within a specific radius, total charging points available at nearby stations, proximity to high-traffic areas, zoning data, and the population demographic and median income of the zone in which the parking lot is located.
Standardization and Weights
Each parking lot was assigned a percentile rank (0-1) for each feature relative to all other locations. The standardized features were then assigned weights based on their importance to EV charging demand. For instance, the "chargers in radius" feature received a negative weight, as we aim to recommend parking lots that are farther from existing stations. On the other hand, features that likely indicate higher demand, such as "proximity to high-traffic areas," were given positive weights.
Demand Score
After assigning the weights, we calculated a demand score for each parking lot. To normalize the scores, we applied Min-Max scaling, ensuring that the absolute value of all the scores fall within a range of 0 to 1. Higher scores indicate a stronger recommendation, while negative scores suggest that the parking lot should not be recommended.
Recommendation Method
The demand scores for the parking lots were ranked from highest to lowest and assigned one of three recommendation labels: highly recommended, recommended, or not recommended. The parking lots are then displayed on the interactive map along with additional details about each parking lot.
Results
We developed an interface that allows users to input a zip code and specify the
number of parking lots they want recommended. These
recommendations are displayed
on an interactive map, with a tooltip feature to display additional details for
each parking lot. We also
included a factor analysis section at the bottom,
explaining how the recommendations were calculated and why specific locations
were
suggested. Overall, the model performs well, recommending parking lots based on user inputs
and categorizing them as highly recommended,
recommended, or not recommended. The model returns the
maximum feasible locations if the user requests more recommendations
than are available in a given zip code.
After analyzing the results, we identified several trends with the top-ranked
locations. Many were distant from existing infrastructure,
indicating the tool's
effectiveness in identifying underserved areas. Other trends include proximity to
mixed-use zones, commercial
districts, high-traffic areas, and high population
density, showing a need for chargers in more populous areas.
However, the model does have a few limitations. The parking lot dataset relies on
user-contributed data, which may lead to missing
or incomplete information.
Additionally, the tool is currently limited to SDG&E's service areas and uses
static assumptions, restricting
its generalizability to other regions and future
urban development. Furthermore, the model hasn't been tested with city planners or EV
drivers, which could limit the applicability of the recommendations.
Conclusion
Our project addresses the growing need for more EV charging stations
to support the increased demand for electric vehicles. By utilizing
our interactive map, we aim to identify areas with the greatest need
for new chargers, helping assist SDG&E's clean transportation team
in planning the placement of future EV chargers. While there are some
limitations to our work, we hope this project can be used as a
foundation
for future expansion and improvement in EV infrastructure.
Ultimately, this project is a step toward meeting the rising demand
for electric vehicles, supporting the transition to clean transportation,
and contributing to a more sustainable future. We hope to work with city
planners, utility companies, and other stakeholders to
expand EV charging access.