Probabilistic Prediction of Energy Demand and Driving Range for Electric Vehicles With Federated Learning

Today's drivers of battery electric vehicles must deal with limited driving range in a sparse charging infrastructure. An accurate prediction energy demand and is therefore important enables reliable routing charge planning applications. Predictions entail uncertainty, which can be considered directly the use probabilistic algorithms. Machine learning algorithms are frequently applied this context, but data used to train these often distributed over fleet connected vehicles. Federated setting, predictive uncertainty typically not considered. We apply an extension federated averaging algorithm learn neural networks linear regression models communication-efficient privacy-preserving manner. demonstrate performance advantage deterministic using proper scoring rules. Furthermore, we show that improve standard, driver-individual learning. Using predictions, variable safety margins based on destination attainability applied, leading increased effective reduced travel time.