The expansion of renewable energy systems, direct current (DC) microgrids, and the proliferation of electric vehicles (EVs) will significantly elevate the need for bi-directional converters. Exact control techniques are crucial to guarantee excellent performance and enhanced efficiency of these converters. This research presents a deep neural network (DNN)-based controller aimed at accurately regulating bi-directional converters for vehicle-to-vehicle (V2V) and vehicle-to-grid (V2G) applications. This control technique enables the converter to rapidly achieve fresh reference values, hence improving performance and efficiency by markedly decreasing the time of overshoot. Large synthetic datasets are utilised to train the DNN controller by conducting simulations under diverse settings, with results validated by a hardware setup. The real-time efficacy of the DNN controller is juxtaposed with that of a traditional proportional-integral (PI) controller, utilising simulated outcomes from MATLAB Simulink (version 2023a) and a real-time configuration. The converter achieves a new benchmark of around 975 µs using the proposed control method. Conversely, the PI controller requires around 220 ms, indicating that the suggested control strategy significantly outperforms the one employed by the PI controller.

non-isolated bi-directional converter (NIBC); V2V charger; deep learning; high voltage; low voltage