A BP Neural Network Based Optimization Strategy to Suppress the Vibration of Power Electronic Transformer under DAB

The vibration of Power Electronic Transformer(PET) is the result of the strong coupling of electric, magnetic and mechanical fields, showing highly-nonlinear characteristics. At present, there is no suitable vibration suppression strategy for PET. Under dual active bridge(DAB) topology, this paper proposes a novel method to predict the PET vibration by using BP neural network with double hidden layers. Experimental results validate that this method has high accuracy for vibration fitting of PET. Based on this neural network, the phase-shift angle corresponding to minimum vibration is found in the triple phase-shift(TPS) mode of DAB. Compared with common SPS, EPS and DPS at the same transmission power level, the optimized method could reduce the vibration amplitude at the key frequency by 24.06dB, 23.32dB and 21.11dB, respectively.