基于FastICA-SVM的光伏并网逆变器故障诊断

Fault diagnosis of photovoltaic grid-tied inverters based on FastICA-SVM

为解决电力电子器件在发生开路故障时,光伏并网发电系统中的逆变器虽能工作,但存在故障隐患的问题,提出一种基于快速独立成分分析(Fast Independent Component Analysis,FastICA)算法和支持向量机(Support Vector Machine,SVM)相结合的故障诊断方法。首先,采用FastICA对三相电流进行独立成分分离,当得到正弦半波和近似正弦波两个独立成分时,判定逆变器发生单管开路故障;其次,求取旋转电流I_(d)时的频域特征值,将其作为SVM模型的输入值;最后,以不同内积核函数构建SVM,将提取的特征值作为SVM的输入值,逆变器工作状态编码作为SVM的预测输出值,以此进行SVM模型训练。实验结果表明:FastICA能够实现逆变器单管开路故障判定;由二次内积核函数构建的SVM模型故障定位准确率最高,可达97%以上,证明了所提方法的有效性。

In order to solve the problem that the inverters in grid-tied photovoltaic power generation system can still work when the power electronic devices have an open circuit fault,a fault diagnosis method based on the combination of fast independent component analysis(FastICA)algorithm and support vector machine(SVM)is proposed.First of all,the FastICA is used to separate the independent components of three-phase current,and the single power electronic device open-circuit fault is judged when two independent components,sinusoidal half wave and approximate sinusoidal wave,are obtained.Secondly,the eigenvalues of rotating current I_(d) in frequency domain is obtained as input of SVM model.Finally,SVM is constructed with different inner product kernel functions,and the extracted eigenvalue is taken as the input value of SVM,and the working state code of inverter is taken as the predictive output value of SVM.The experimental results show that FastICA can realize the single power electronic device open-circuit fault diagnosis,and the SVM model based on the quadratic inner product kernel has the highest fault location accuracy,reaching above 97%,which proves the effectiveness of the proposed method.