交直流变换器是串联式混合动力汽车（series hybrid electric vehicles ,SHEV）电气驱动系统中实现功率变换以及调速调频的核心装置，由于存在大量功率器件，导致其容易发生各种故障，且受功率以及外界环境等各种因素影响导致故障信号特征提取困难，故障诊断难以准确定位故障发生位置。本文主要针对功率器件发生断路故障问题，基于建立的SHEV电气系统的交直流变换器仿真模型，选择直流侧母线输出电流为特征量，并且充分考虑功率器件的位置，对功率器件开路故障类型进行详细分类，然后对各种故障下直流侧母线输出电流的特性进行分析。利用快速傅里叶变换将故障信号分解到不同的频率段，通过分析比较选取30kHz（k=1，2，3…）频率段信号为故障诊断特征向量，再结合基于遗传算法的BP神经网络实现故障类型识别。仿真结果表明，这种方法可对交直流变换器的开路故障进行有效诊断和识别，预测结果误差非常小，具有计算简便，准确性高的优点。
英文摘要:
AC/DC converter is the core device of series hybrid electric vehicles, which can realize power conversion and frequency modulation. Due to the existence of a large number of power devices, many faults are easy to happen. Besides, because of the effects of power and the external environment, it is difficult to extract fault signal features. So, these reasons make it more difficult to locate fault position. To solve these problems, based on the AC/DC converter simulation model of SHEV electrical system, this paper selects the output current of DC side bus. And considering the location and the number of the power devices, the open-circuit fault types are classified in detail. Then this paper analyzes the characteristics of the output DC current under various faults; and using Fast Fourier Transform, the fault signals are decomposed into different frequency segments. The 30kHz(k=1,2,3…) frequency band signals are selected as fault feature vectors. Finally, BP neural network based on genetic algorithm is used to realize fault type recognition. The simulation results show that this method can effectively diagnose and identify the open circuit faults of AC/DC converters, and the error of the prediction result is very small. Above all, this method has the advantages of simplicity and high accuracy.