Artificial Intelligence-Based Modeling and Control Strategy for Dual Active Bridge Converter With Triple Phase Shift Modulation

Dual active bridge (DAB) converter, known for its exceptional performance in bidirectional power transmission, has emerged as a crucial power conversion unit in energy storage systems and dc power systems. To improve the efficiency under wide voltage variations, the triple phase shift (TPS) modulation is typically used. However, the existing analytical models and control strategies of DAB with TPS modulation mostly depend on complex theoretical derivations, neglecting nonideal factors or exhibit heavy computational burdens. To tackle this issue, an artificial intelligence (AI)-based modeling and control strategy is proposed. First, a numerical simulation model incorporating the dynamic behavior of DAB circuit during dead time and parasitic parameters is developed to efficiently generate the training dataset. Building upon this, a general AI model of TPS modulation is precisely formulated, which is concise and applicable to various circuit parameters and all switching modes. Subsequently, a multilayer heuristic algorithm is employed to effectively derive optimal steady-state strategies. In addition, a concise artificial neural network is constructed to fit the offline control law and generate the optimal strategies online. Furthermore, an adaptive neural network sliding mode controller is designed. It integrates the general AI model with nonlinear sliding mode control theory, while a recurrent neural network compensates for errors adaptively, exhibiting excellent robustness and transient performance. Finally, comparative experiments demonstrate the effectiveness and superiority of the proposed approach.