Current Stress Optimization Method for Dual Active Bridge Converter Based on TabNet-LN-LSTM Co-prediction and Particle Swarm Optimization

The dual active bridge（DAB） converter has gained widespread attention in many industrial applications due to its excellent power density and bidirectional power transfer capability. With the increasing requirements for energy efficiency and reliability of power electronic devices, the current stress of the DAB converter has become one of the key performance metrics. Excessive current stress not only increases power device losses and reduces system efficiency, but also adversely affects converter reliability and operational lifespan. To address the above problems, a current stress optimization method based on Tab Net-LN-LSTM co-prediction and particle swarm optimization is proposed. The method uses TabNet and long-short term memory neural network with layer normalization（LN-LSTM） to build the inductor current time-series prediction model and combines the particle swarm optimization（PSO） algorithm to optimize the current stress of the DAB converter under different operating conditions. The algorithmic and hardware experiments demonstrate that the proposed method can not only accurately predict the inductor current waveforms with an average absolute error of 0.352 5 and a coefficient of determination as high as 97.17% compared with the measured hardware waveforms, but also effectively reduce the current stress of the DAB converter, thus improving the overall efficiency and reliability of the system.