Privacy Preserving Charge Location Prediction for Electric Vehicles

TL;DR

This work tackles privacy in predicting EVs' next charging locations by introducing a Federated Learning Transformer Network (FLTN) that trains on-device and shares only weights. It enhances privacy through peer-to-peer weight sharing and augmentation among non-transitory EVs, with community DERMS performing weight aggregation via FedAVG and redistributing the global model for continual learning. Empirical results on a Chicago-based EV taxi mobility dataset show FLTN achieves about 92% accuracy, slightly below a centralized 98% baseline, while providing stronger privacy guarantees and resilience to data leakage and model inversion. The approach offers a scalable, privacy-centric pathway for energy providers to forecast demand and manage DERMS across dynamic EV networks in urban environments.

Abstract

By 2050, electric vehicles (EVs) are projected to account for 70% of global vehicle sales. While EVs provide environmental benefits, they also pose challenges for energy generation, grid infrastructure, and data privacy. Current research on EV routing and charge management often overlooks privacy when predicting energy demands, leaving sensitive mobility data vulnerable. To address this, we developed a Federated Learning Transformer Network (FLTN) to predict EVs' next charge location with enhanced privacy measures. Each EV operates as a client, training an onboard FLTN model that shares only model weights, not raw data with a community-based Distributed Energy Resource Management System (DERMS), which aggregates them into a community global model. To further enhance privacy, non-transitory EVs use peer-to-peer weight sharing and augmentation within their community, obfuscating individual contributions and improving model accuracy. Community DERMS global model weights are then redistributed to EVs for continuous training. Our FLTN approach achieved up to 92% accuracy while preserving data privacy, compared to our baseline centralised model, which achieved 98% accuracy with no data privacy. Simulations conducted across diverse charge levels confirm the FLTN's ability to forecast energy demands over extended periods. We present a privacy-focused solution for forecasting EV charge location prediction, effectively mitigating data leakage risks.

Figures (8)

• Figure 1: The pipeline diagram above consists of processes and stages for our FLTN system from data input to our final output EV next charge location prediction.
• Figure 2: Chicago city map - community areas light yellow. The top 5 communities for taxi activity in green and orange are 32, 6, 28, 7, and 8 which has the highest density of taxi activity in orange.
• Figure 3: Centralised modelling results. Three machine learning architectures; Transformer, BiLSTM, and CNN. Models used the same communities as our decentralised solution using groups of 100, 200, and 300 EVs to understand which architecture was best suited to our problem area.
• Figure 4: This figure represents Chicago city communities and the density of taxi trips per community covering 12 months. The X-axis represents community IDs, the Y-axis density of taxi trips, highest density communities are represented by orange and green, which also represent the CBD for Chicago City.
• Figure 5: This figure illustrates a sample of 500 EVs over a 12-month period, differentiating transitory and non-transitory EV taxi trips based on pickup and dropoff within the same community.