双有源桥变换器开路故障的精细化建模与在线诊断方法

The dual-active-bridge (DAB) converter is a popular dc–dc converter topology for high-power applications. An open-circuit fault (OCF) in a DAB converter can produce a dc bias in its transformer currents, which can saturate the transformer, resulting in significant power losses and power quality deterioration. In this article, a refined modeling and an online OCF diagnosis method are proposed for OCF in DAB converters. First, the causes of voltage and current waveform distortion caused by OCF are analyzed in detail. The reasons for differences in transformer voltage and current waveforms due to different OCF positions are studied. The importance of critical parameters such as the magnetizing inductance of transformers and parasitic capacitance of power switches to the analysis of circuit operation characteristics under OCF conditions is clarified. Based on this theoretical research, a refined DAB converter modeling incorporating critical parameters is proposed. In contrast to current models, this proposed model offers improved accuracy in depicting variations of primary-side electrical parameters under OCF conditions. In addition, an OCF diagnosis method is proposed. This method utilizes the cycle-average value of the transformer primary-side current for OCF detection and localization. Unlike prior methods, the proposed method does not necessitate supplementary hardware sensing conditions, nor does it require the converter to operate in a specific switching state. Finally, the online deployment of the OCF diagnosis method based on an FPGA is completed. Experimental results verify that the proposed model effectively simulates variations in the primary-side and secondary-side currents of the transformer, as well as the magnetizing current, under OCF conditions. As a fault feature, the mean absolute error (MAE) of the primary-side current is within 0.5 A and the period average error is within 0.2%. Moreover, the online OCF diagnosis method enables OCF detection within 1 switching cycle and OCF localization within 6 switching cycles.