Machine Learning Model for EV Charging Station Placement Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for predicting popularity and visitation metrics for locations without electronic vehicle charging stations are inaccurate due to incomplete data and lack of statistical sampling, making it challenging to determine optimal placement and usage of EVCSs.
Innovation Solution
Training a machine learning model using visitation data and popularity metrics from existing EVCS locations to predict metrics for new locations, including visitation lift calculations for EVCS placement optimization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If statistical sampling from mobile phones and cars is used to collect POI visitation data, then data collection coverage is improved, but measurement precision deteriorates due to large discrepancies with actual visitation
Solution Approach 1:
The patent introduces machine learning models as intermediaries that bridge the gap between statistical sampling data and actual visitation metrics. The models learn patterns from locations with existing EVCS data and apply these patterns to predict metrics for locations without EVCS, thereby mediating between incomplete sampling data and accurate visitation predictions
Solution Approach 2:
The patent creates predictive copies of visitation metrics for locations without EVCS by training on data from locations with EVCS. The machine learning model generates synthetic visitation data for new locations based on patterns learned from existing locations, effectively copying the relationship between location characteristics and visitation metrics
2Measurement precision
If data is collected only from locations with existing EVCS, then measurement precision is improved, but adaptability deteriorates because no metrics are available for locations without EVCS
Solution Approach 1:
The patent creates a universal machine learning model that serves multiple functions: it accurately predicts metrics for locations with existing EVCS while also extending predictive capability to locations without EVCS. The model is trained on diverse features including location characteristics, POI types, and regional attributes, enabling it to generalize across different location types and EVCS deployment scenarios
Solution Approach 2:
The patent performs preliminary training of the machine learning model using data from locations with existing EVCS before applying it to predict metrics for locations without EVCS. This preliminary action of learning from available data enables the system to subsequently predict metrics for new locations where no EVCS data exists yet
3Measurement precision
If machine learning models are trained on existing EVCS data, then prediction accuracy for new locations is improved, but device complexity increases
Solution Approach 1:
The patent segments the prediction task into distinct components: location feature extraction, POI type classification, and visitation metric prediction. The machine learning model processes different input features (location characteristics, regional attributes, POI types) separately and combines them to generate predictions, making the complex prediction task more manageable and interpretable
Data Source
AI summary
Techniques are described herein for predicting popularity metrics and/or visitation metrics that are used in the selection of a point of interest (POI) for placement of an electric vehicle charging station (EVCS). The techniques involve training a machine learning model based on information obtained about POIs at which EVCSs are already installed. The information used to train the machine learning model includes, for each existing installation location: (a) visitation data that describes visitation features, and (b) popularity metrics and/or visitation metrics that have been generated for the location. When the machine learning model has been trained, the trained machine learning model predicts popularity metrics and/or visitation metrics for a POI location at which no EVCS has been installed based on the visitation data of that POI.