p27 / Conference
Federated Learning for Electric Vehicle Energy Demand Prediction Across Distributed Users
Availability, Reliability and Security: ARES 2026 EU Projects Symposium Workshops, LNCS 16903, pp. 166-184 · 2027 · Online 2026
In plain language
Daily EV energy demand is forecast across heterogeneous users and fleets without requiring central collection of their raw records.
Research context
The work connects demand forecasting with deployable collaboration across EV data owners.
Approach
A shared vehicle-day preprocessing pipeline combines state of charge, distance, history, calendar, and weather. Local, centralized, and federated models are compared, with asynchronous FL tested on Raspberry Pi devices.
Findings
Distributed models approach centralized prediction quality, and buffered asynchronous training is practical in the reported heterogeneous testbed.
Summary based on abstract. Publisher citation year is 2027. First online publication was 24 August 2026, during ARES 2026.