基于盲源分离和机器学习的光伏并网逆变器故障诊断

Fault Diagnosis of Grid-Connected PV Inverters Based on Blind Source Separation and Machine Learning

针对光伏并网逆变器单个开关管发生开路故障不易察觉的问题,提出了一种基于盲源分离和机器学习的诊断方法。首先,采用FastICA算法实现单个开关管开路故障的判定;其次,提取旋转电流在时域和频域下的特征值;最后,以旋转电流特征值为输入、逆变器工作状态编码为输出进行机器学习模型训练,并对模型进行交叉验证。仿真实验结果表明,该方法的开路故障诊断准确率较高。

Aiming at the problem that the open circuit fault of a single switching tube in grid-connected photovoltaic inverters is not easily detected,a diagnosis method based on blind source separation and machine learning was proposed.Firstly,FastICA algorithm is used to determine the open-circuit fault of a single switch tube.Secondly,the eigenvalues of rotating current in time domain and frequency domain are extracted.Finally,the machine learning model is trained with the characteristic value of rotating current as input and inverter operating state coding as output,and the model is cross-verified.Simulation results show that the open circuit fault diagnosis accuracy of this method can reach more than 98.9%,which has high application value.