【作者】 王荣杰；

【导师】 胡清；

【作者基本信息】 广东工业大学 ， 电力电子与电力传动， 2006， 硕士

【摘要】 随着电力电子技术的迅猛发展,实现能量变换的电力电子整流装置,由于其效率高、控制灵活方便、易实现等优点,使其的应用日益广泛,同时电力电子整流装置的故障问题也越来越突出,因此在电力电子整流装置中应用自动故障诊断技术,是有其现实意义和经济意义的,开展相关的理论和方法研究尤为重要。 课题以发展和完善电力电子整流装置故障的智能诊断方法为目的,针对电力电子整流装置故障诊断中的故障特征提取和识别两个关键技术问题进行了深入研究和分析,主要完成了如下的研究工作: 提出一种基于PCA-神经网络的电力电子整流装置故障诊断方法。首先对故障信号用主元分析法(PCA)提取特征向量,然后用神经网络进行训练和测试。通过三相可控整流电路晶闸管断路故障诊断实验结果表明,该方法能够简化神经网络的结构,提高网络的训练速度,并获得了很好的诊断效果。 提出利用小波包分析理论对电力电子整流装置的故障信号进行特征提取,提取出故障信号的能量特征作为故障分类器的输入向量,由故障分类器对各种故障进行识别和诊断。通过十二脉波可控整流电路晶闸管断路故障诊断仿真结果验证了该方法的有效性。 提出一种基于SVM的电力电子整流装置故障诊断方法。研究一种改进的多故障SVM分类算法,基于这种算法的思想提出基于SVM的电力电子整流装置故障诊断方法。通过十二脉波可控整流电路晶闸管断路故障诊断实验结果表明,该方法能够相当准确地对故障进行诊断,诊断精度高;与BP神经网络进行比较,该方法计算效率高,且在小样本下具有很好的推广能力,在解决电力电子整流装置故障问题上有着很好的实用价值和应用前景。更多还原

【Abstract】 With rapid development of power electronics technology, power electronics rectifier of realizing energy conversion, because of its efficient, flexible and convenient control, easy to achieve and so on, make its application widespread increasingly. The fault problems of the power electronics rectifier are more and more prominent at the same time, therefore applies the automatic fault diagnosis technology in the power electronics rectifier has its practical significance and the economical significance, it is particularly important that the relevant theoretical and methodological research.The topic takes the development and improvement of power electronics rectifier fault intelligent diagnosis method as a goal. Against features extraction and diagnosis of two key technology in fault diagnosis for power electronics rectifier has conducted further research and analysis. The main research work completed as follows:Fault diagnosis method for power electronics rectifier based on PCA-Neural Network was proposed. First extract the feature vector from the fault signal with the principal component analytic (PCA) method, and then use neural network training and testing. Experimental result of thyristor open circuit fault diagnosis in power electronics rectifier showed that this method can simplify the structure of the neural network, improve the training speed of the network, have obtained very good diagnostic effect.Putting forward method that using wavelet packet analysis theory to carry on the characteristic from the fault signal of power electronics rectifier, extract its energy as input vector of fault classifier, carries on the classification and the diagnosis by the fault classifier to each kind of fault. Has confirmed this method validity through twelve pulse controlled rectifier thyristor open circuit fault diagnosis simulation result.Proposes fault diagnosis method for power electronics rectifier based on the SVM. An improved multi-fault SVM classification algorithm were studied, put forward fault diagnosis method for power electronics rectifier based on the SVM with ideal of the algorithm. Indicated through twelve pulse controlled rectifier thyristor open circuitfault diagnosis experimental result, this method can quite accurately diagnosis fault, and the diagnostic precision is high;Comparison with BP neural network, this method efficient and have a good outreach capacity under small-sample. It has the very good practical value in the solution to fault question for power electronics rectifier and the application prospect.更多还原