EV Charge Point Utilization Prediction via Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The widespread adoption of electric vehicles is hindered by the inadequate and inefficient electric vehicle charging infrastructure, which lacks sufficient charging points, leading to high demand, broken chargers, high costs, and consumer confusion, posing a significant barrier to EV adoption.
Innovation Solution
A method utilizing geospatial data processing and supervised machine learning to predict electric vehicle charge point utilization, identifying optimal locations for new charging points based on static and dynamic map features, such as POI proximity and traffic density, to improve return-on-investment and infrastructure efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If EV charging infrastructure is expanded to meet growing demand, then charging availability improves, but investment costs increase substantially
Solution Approach 1:
The patent applies local quality by transitioning from uniform geographic dispersion of charging stations to location-specific deployment based on predicted utilization. Machine learning models analyze static map features (POI categories, road functional class, population density) and dynamic features (traffic density, weather, time of day) to identify high-demand locations, ensuring charging infrastructure is concentrated where it will be most utilized rather than distributed equally across all areas.
Solution Approach 2:
The patent employs preliminary action by using machine learning models to predict future charging utilization at candidate locations before infrastructure is built. The system trains models on historical charge point utilization data and static/dynamic map features to forecast which locations will have highest demand, allowing planners to proactively place charging stations at optimal locations before deployment, maximizing expected utilization from the outset.
2Area of stationary object
If charging stations are placed in EV charge point deserts or equally dispersed, then geographic coverage improves, but utilization efficiency decreases
Solution Approach 1:
The patent applies parameter changes by shifting the decision criteria for charging station placement from uniform geographic parameters to utilization-prediction parameters. Instead of placing stations based on equal spatial intervals or targeting only underserved 'charge point desert' areas, the system uses machine learning models that process static map features (POI proximity, road class, population density) and dynamic features (traffic density, weather, time) to predict utilization, fundamentally changing how placement locations are selected to maximize both coverage and efficiency.
3Adaptability or versatility
If more charging points are deployed to reduce consumer barriers, then EV adoption accelerates, but infrastructure complexity increases
Solution Approach 1:
The patent applies feedback by implementing a closed-loop system where machine learning models continuously predict charging utilization at candidate locations, and these predictions feed back into the deployment decision-making process. The system processes static map features and dynamic features to generate utilization forecasts, which then inform where new charging stations should be placed, creating a feedback-driven approach that adapts infrastructure deployment to actual and predicted demand patterns, simplifying management by focusing resources on high-utilization locations.
Data Source
AI summary
Embodiments described herein relate to predicting the utilization of electric vehicle (EV) charge points. Methods may include: receiving an indication of a plurality of candidate locations for EV charge points; determining static map features of the plurality of candidate locations; inputting the plurality of candidate locations and static map features into a machine learning model, where the machine learning model is trained on existing EV charge point locations, existing EV charge point static map features, and existing EV charge point utilization; determining, based on the machine learning model, a predicted utilization of an EV charge point at the plurality of candidate locations; and generating a representation of a map including the plurality of candidate locations, where candidate locations of the plurality of candidate locations are visually distinguished based on a respective predicted utilization of an EV charge point at the candidate locations.