交直流混合微电网中多单元逆变器故障诊断

Fault diagnosis of multi⁃unit inverter in AC/DC hybrid microgrid

为提高交直流混合微电网中交流子网上分布式发电单元多个单元逆变器同时发生故障时的诊断识别率,提出一种基于深度学习的交直流混合微电网逆变器中IGBT开路故障的诊断方法。即提取交直流混合微电网中交流子网上分布式电源公共连接点的原始信号进行数据预处理;之后再进行小波包分解,获得信号不同频次的特征数据;再将得到的特征数据进行处理,使用GASF(格莱姆角求和矩阵)转化为二维图片;最后输入到ResNet18网络中进行学习训练,实现交直流混合微电网中逆变器开路故障的诊断。通过实验验证,该方法可以区分交直流混合微电网不同发电单元及多个发电单元的逆变器中IGBT开路故障,准确度可达93%。与其他诊断方法进行对比,该方法诊断准确率高,证明其有效性。

In order to improve the diagnosis and recognition rate of multi⁃unit inverters of distributed generation unit in AC subnet in AC/DC hybrid microgrid,a deep learning based diagnosis method for IGBT open circuit fault in AC/DC hybrid microgrid inverter is proposed.The original signal of the common connection point of distributed power in the AC subnet in the AC/DC hybrid microgrid is extracted for data preprocessing,and then the wavelet packet decomposition(WPD)is implemented to obtain the characteristic data of the signal at different frequencies.The obtained characteristic data is processed and converted into two⁃dimensional pictures by GASF(Gramian angular summation field),and then input into ResNet18 network for learning and training,so as to realize the open circuit fault diagnosis of inverter in AC/DC hybrid microgrid.The experiments show that the proposed method can distinguish IGBT open circuit faults in inverters of different power generation units and multiple power generation units in AC/DC hybrid microgrid,and its accuracy can reach 93%.In comparison with other diagnostic methods,this method has high diagnostic accuracy,which has verified its effectiveness.