EV-Planner: a machine learning approach to electric vehicle charging infrastructure planning

Area of Science:

• Sustainable Transportation
• Operations Research
• Computer Science

Background:

• Equitable adoption of electric vehicles (EVs) hinges on robust public charging infrastructure.
• Current planning methods often lack comprehensive criteria for optimal station placement.

Purpose of the Study:

• To introduce EV-Planner, a novel software tool for strategic EV charging station placement.
• To address limitations in existing approaches by incorporating multiple optimization objectives.

Main Methods:

• Formulating EV charging station placement as a constrained multi-objective optimization problem.
• Employing a clustering-based approximation with a Suggest-Accept (SA) criterion.
• Utilizing a neural network for hyperparameter tuning within the SA clustering process.

Main Results:

• EV-Planner reduced average user-to-station distances by 52.3%.
• Decreased the number of users over three miles from a station by up to 10.7 times.
• Achieved a 71.7% reduction in overloaded EV stations compared to baseline methods.

Conclusions:

• EV-Planner offers a significant advancement in planning equitable and efficient EV charging infrastructure.
• The tool demonstrates substantial improvements in accessibility, coverage fairness, and load balancing.
• Provides a promising solution for facilitating widespread electric vehicle adoption.