A PSO-ANN-aided DC-bias Current Suppression Strategy for Three-phase DAB converter with Variable Duty Cycles

The dc bias in inductor current of three phase Dual active bridge (3ph-DAB) converter presents notable risks, such as overcurrent stress and increased power loss. Additionally, the intricate structure and three control variables in variables duty cycles (VDC) modulation pose significant challenges in analysis and deduction procedures, which suffers from high manpower burden and low accuracy. In this context, this paper proposes an Artificial Neural Network (ANN) based dc bias suppression method. Firstly, the traditional method is reviewed by applying space vector and piecewise modeling method. Furthermore, the optimization objective function is established by analyzing the behavior of the converter, focusing on the deviation of the peak value of the inductor current between the transient and subsequent steady state. The PLECS software is utilized to acquire optimal datasets of transient variables that minimize the objective function via particle swarm optimization (PSO) algorithm. Subsequently, two ANNs are then developed utilizing these optimal datasets, yielding data-driven models for transient variables. Finally, experimental results are given to validate the effectiveness of the proposed method, showing a reduction in the transient period to within one-third of the switching period.