Machine Learning-Based EV Charging Management System for Vehicle-to-Home and Vehicle-to-Building

With the increase in adoption of electric vehicles (EVs) around the world, there has been an increasing interest in bidirectional EV charging that allows EV to act as an energy storage device and thus, provide energy arbitrage in smart home and building through vehicle-to-home (V2H) and vehicle-to-building (V2B) setup. However, a smart energy management system is required to efficiently manage the exchange of energy between EVs and homes and buildings in order to fully capitalise on the potential of V2H and V2B. In this article, a machine learning (ML) based approach that uses long short-term memory (LSTM) to predict EV charging/discharging power in real time has been proposed. The proposed approach factors various parameters, such as the state of charge (SoC) of the EV, generation from rooftop solar PV and time of use electricity tariff to predict EV charging/discharging power. The proposed model is trained using historical data and then deployed to predict EV charging/discharging power in V2H and V2B system. The proposed LSTM based algorithm is comprehensively tested using real life data and the results demonstrate that the proposed algorithm provides accurate predictions without violating the physical constraints associated with EVs.