Design and Analysis of Digital Twin Models for Dual Active Bridge

Digital twins (DTs) are emerging as effective tools for power electronic converters, addressing important objectives such as real-time monitoring, fault detection and predictive maintenance. Among these power converters, the dual active bridge (DAB) stands out for DC-DC conversion applications. However, the non-linear dynamics of DABs pose considerable challenges to their accurate modeling. To address these challenges, this paper investigates the use of machine learning (ML) and deep learning (DL) models for DT implementation in DABs. Specifically, XGBoost and dense neural network (DNN) models are employed to construct DTs compatible with low-cost microcontrollers. These DTs are operated in parallel with the physical DAB converter to predict its output voltage in real-time. Experimental results demonstrate that both of the models perform well in terms of accuracy and applicability to low-cost microcontroller. However, the DNN is more reliable and the XGBoost is simpler.