Block-FeDL: Electric Vehicle Charging Load Forecasting using Federated Learning and Blockchain

Abstract

The increased charging demand resulting from the rapid development of electric vehicles (EVs) poses various challenges to the stable operation of the distribution network and smart grid. Due to the stochastic EV charging behaviour, the high charging demand at the charging stations (CSs) elevates the load curve which may lead to a spatially imbalanced load demand. As such, forecasting the highly stochastic EV charging load considering an individual EV's unique charging behaviour can result in maintaining the safe operation of the grid and distribution network. Therefore, in this work, we propose Block-FeDL, a blockchain-based Federated Learning (FL) approach for EV charging load forecasting considering the private and sensitive charging information of each EV user. Thereafter, we use a Bidirectional Long Short Term Memory (BiLSTM) model within the FeDL for predicting the EV charging load patterns at each client. Moreover, instead of using a centralized server for global model aggregation, we use blockchain technology, where the model aggregation is performed in a decentralized manner and the local model parameters shared by the FL clients can be validated and securely recorded. Lastly, the results show that the Block-FeDL outperform the second-best baseline method by 95\%, 96\% and 77\% in terms of mean square error (MSE), mean absolute error (MAE), and root mean square error (RMSE) for forecasting the EV charging load.

Tasks

Reproductions