Our model is based on machine learning, an artificial intelligence technology that learns from real-world data to make reliable forecasts.
We trained the model on more than 100,000 charging points across different types of locations (car parks, hotels, retail areas, highways, etc.). This allows it to recognize energy consumption patterns depending on the context.
The model combines several types of information:
- Traffic data (volume, road types, peak hours, weekday vs weekend trends), from MyTraffic’s Traffic datas,
- Local environment (presence of shops, restaurants, service stations, competing charging stations), from points of interest and infrastructure datasets,
- Socio-demographics (population density, purchasing power, local EV adoption rate), from public statistics and demographic studies,
- Visitor behavior (average dwell time, attractiveness of nearby destinations), from MyTraffic’s Traffic datas.
All of this data is processed using a Random Forest algorithm. This technique aggregates the results of hundreds of decision trees (different possible scenarios) to deliver a more robust and accurate prediction.
This approach highlights the key drivers of consumption, which vary depending on the location type: for example, highway traffic is a strong predictor on service areas, while in urban zones, local EV ownership and purchasing power play a larger role.
In practice, the model delivers energy consumption forecasts with a median error rate between 11% and 20%, depending on the charging speed – providing reliable, actionable insights for decision-making.
Comments
0 comments
Please sign in to leave a comment.