Enhancing Smart Grid Efficiency: Blockchain-Based Federated Learning for EV Energy Forecasting

The growing popularity of electric vehicles (EVs) is driven by their advantages over conventional fuel-powered cars. However, their integration into the power grid presents challenges such as increased energy consumption and peak load management. This study proposes a blockchain-enabled federated learning (BCFL) framework that utilizes linear regression techniques to enhance EV energy demand prediction. The data collected from EVs is securely stored on a blockchain network, ensuring restricted access through encryption mechanisms. A federated learning approach allows each EV to train a localized model without sharing raw data, preserving privacy. The trained model parameters are aggregated and shared via blockchain, ensuring data integrity and security. This approach is unique in its evaluation of BCFL’s communication overhead and latency issues, offering an optimized strategy to reduce delays and improve system efficiency. Implementation results validate the accuracy of the proposed framework in forecasting EV energy consumption. A real-world dataset consisting of over 60,000 EV charging transactions from Boulder City, Colorado, was used to train the models. The findings confirm the reliability of the system, as all models achieved R² values exceeding 0.91, demonstrating high precision in energy