Full-Parameter Identification of Buck Converter Through BP-NN Fitting Explicit Time-Domain Relationships
Zhennan She, Yisi Liu, Wen Cao, Guipeng Chen
Abstract
In this paper, back propagation neural network (BP-NN) is employed to identify all component parameters of Buck converter by fitting the explicit time-domain relationships. Thanks to the powerful ability of fitting the non-linear relationship of BP-NN, the proposed method can effectively avoid a large amount of direct calculation and thus it is convenient to implement. Meanwhile, the explicit time-domain relationships between voltage/current and component parameters in the Buck converter are revealed and utilized to train BP-NN. Hence, higher parameter identification accuracy and lower network configuration requirements are favorably achieved in comparison with the conventional neural network based method with the ambiguous relationship. Besides, the proposed method is also experimentally validated on a closedloop Buck converter. It is not only workable under different operating conditions, but also is effective to monitor parameter variation during the component aging process