Data-Driven Power Sharing Control Within Multisource Paralleled DAB Converters

To achieve redundancy requirements for transportation electrification, a very commonly used method is to have multiple energy storage sources connected in parallel and feed a common dc bus through dc/dc power converters. The dual-active bridge (DAB) converter, in that respect, is an appealing topology to interface these energy storage elements with the main dc grid due to its simple structure and high efficiency. However, conventional optimum control has limitations in the power sharing control scheme. To address these challenges, this article proposes a data-driven method using the artificial neural network (ANN) to generate the optimal control solutions and achieve desired current sharing among parallel DABs. The design process begins with data generation from extensive DAB simulations. Subsequently, the derived data is processed and clustered, such to map DAB phase shifts with its power output. Then the ANN is used to train a surrogate model to map the power demand and phase shifts. With the aid of ANN model, the small-signal model of the DAB converter is also simplified. Finally, the proposed method is implemented in a three-DAB system. The current sharing performance, efficiency, and the small-signal model have been proved by experiments.