Improving Supercharger Forecasts
We are maniacal about delivering the best charging experience. This means building a network where you almost never have to wait to charge. For the rare times when a wait does occur, we need to provide the most accurate estimates so you can plan with confidence.
Our Trip Planner intelligently selects Supercharger sites to minimize total travel time. It's powered by a model monitoring real-time traffic in a geofence around Superchargers and predicts how many vehicles intend to charge. These predictions of Supercharger occupancy and queue length optimize your route and help set accurate expectations for your trip.
Supercharger sites are often co-located with amenities, offering convenient stops while you charge. The mixed purpose traffic at these sites makes queue predictions challenging, but we found a fix. We’re now rolling out an updated machine learning model to better identify vehicles that have an intent to charge:
The model is trained on 9 million miles of aggregated and anonymized vehicle trajectory data within the geofence of Superchargers globally.
It reduces queue length estimation error to 20%. That means in the very rare case of 10+ vehicles waiting, we can now predict the expected queue with an error of just 1-2 vehicles.
We are uniquely positioned to deliver this level of charging intelligence through our vertical integration. There is still more work needed to nail these forecasts and we're already working on the next release.
https://x.com/TeslaCharging/status/2047387324573737381