三电平逆变器的决策树SVM故障诊断

发布时间：2018-04-20 09:29

本文选题：逆变器 + 三电平　；参考：《电子测量与仪器学报》2017年02期

【摘要】：针对二极管箝位型三电平逆变器的开路故障诊断问题,提出一种基于决策树支持向量机(decision tree support vector machines,DT-SVM)的故障诊断方法。以逆变状态为例,首先分析逆变器主电路的运行情况并进行故障分类,然后以中、上、下三种桥臂电压为测量信号,采用小波多尺度分解法提取特征信号,进而利用粒子群聚类算法(particle swarm clustering algorithm)生成决策树SVM分类模型,最终实现了三电平逆变器的多模式故障诊断。仿真结果表明,本方法在使用了较少分类模型的情况下完成故障诊断任务,相较于BP神经网络、一对一结构的支持向量机和极端学习机等方法,在10%白噪声扰动下对于三电平逆变器多模式开路故障的诊断精度可达98.46%,算法具有更好的准确性和鲁棒性。
[Abstract]:For the open circuit fault diagnosis of diode clamped three-level inverter, a fault diagnosis method based on decision tree support vector machine (DT-SVM) is proposed. Taking the inverter state as an example, the operation of the main circuit of the inverter is analyzed and the fault classification is carried out. Then the characteristic signals are extracted by wavelet multiscale decomposition with the middle, upper and lower leg voltages as the measuring signals. Furthermore, the particle swarm clustering algorithm is used to generate the decision tree SVM classification model, and finally the multi-mode fault diagnosis of the three-level inverter is realized. The simulation results show that this method can accomplish the fault diagnosis task with fewer classification models, compared with BP neural network, one-to-one structure support vector machine and extreme learning machine, etc. Under 10% white noise disturbance, the accuracy of multi-mode open-circuit fault diagnosis of three-level inverter can reach 98.46. The algorithm has better accuracy and robustness.
【作者单位】：江南大学教育部轻工过程先进控制重点实验室;
【基金】：国家自然科学基金(61203092) 中央高校基本科研业务费专项资金(JUSRP51733B) 江苏省产学研前瞻性联合研究项目(BY2015019-21)资助
【分类号】：TP18;TM464

【相似文献】