基于小波变换信号处理的电力电子电路故障诊断研究

发布时间：2018-03-24 02:11

本文选题：电力电子电路　切入点：故障诊断　出处：《合肥工业大学》2017年硕士论文

【摘要】：随着电力电子技术得到广泛地应用,电力电子设备的结构变得越来越复杂、规模越来越庞大,同时电力电子设备的故障问题也越来越突出。电力电子电路作为电力电子设备的核心组成部件,它的故障将导致电力电子设备甚至整个系统的失效,造成巨大的损失与伤害。为了保证系统安全可靠地运行,应及时地对故障电路进行有效诊断,这就使得电力电子电路故障诊断理论和方法的研究受到人们的越来越多重视。本文对电力电子电路故障诊断中故障特征提取和故障模式辨识的关键问题进行研究,包括以下内容:(1)首先对电力电子电路各故障类型进行分析,建立其各故障形式的仿真电路模型。针对电力电子电路故障特点,考虑实际运行条件下噪声干扰,对所有的故障信息叠加噪声,以模拟实际仿真信号;(2)提出基于小波包变换的方法对故障信号进行故障特征提取,首先采用小波包分解优预测变量阈值法对信号进行消噪预处理,再利用小波包变换能量谱提取原始故障特征向量,结合主成元分析思想,对其进行降维处理,再一次使故障特征得到凸显,得到新的故障特征向量;然后研究改进BP神经网络的故障辨识技术,对故障特征向量进行辨识处理,仿真实例验证方法的有效性,获得较高的诊断率;(3)考虑到小波包故障特征提取复杂程度,为进一步提高故障特征辨识度,提出基于交叉小波变换提取故障特征向量。交叉抗噪特性使其无需进行降噪处理且可以同时对两信号直接分析对比得到信号之间幅值与相位关系,将它们组成原始故障特征向量;结合主成元分析得到最终故障向量;再次将新的故障特征向量输入改进BP神经网络进行故障辨识。对上述故障特征提取方法进行分析对比,仿真实例证明所提方法的有效性且具有更高的诊断率。
[Abstract]:With the wide application of power electronics technology, the structure of power electronic equipment becomes more and more complex, and the scale of power electronic equipment becomes larger and larger. At the same time, the fault problem of power electronic equipment is more and more prominent. Power electronic circuit, as the core component of power electronic equipment, will lead to the failure of power electronic equipment or even the whole system. In order to ensure the safe and reliable operation of the system, the fault circuit should be diagnosed effectively in time. As a result, more and more attention has been paid to the theory and method of power electronic circuit fault diagnosis. In this paper, the key problems of fault feature extraction and fault mode identification in power electronic circuit fault diagnosis are studied. Including the following contents: (1) first of all, the types of power electronic circuit faults are analyzed, and the simulation circuit models of each fault form are established. According to the fault characteristics of power electronic circuits, the noise interference under the actual operating conditions is considered. For all the fault information, the noise is superimposed, and the simulation signal is simulated. (2) A method based on wavelet packet transform is proposed to extract the fault feature of the fault signal. Firstly, the wavelet packet decomposition optimal predictive variable threshold method is used to pre-process the signal noise reduction. Wavelet packet transform energy spectrum is used to extract the original fault feature vector, combined with the main component analysis idea, the dimension reduction is carried out, the fault feature is highlighted again, and a new fault feature vector is obtained. Then the fault identification technology of improved BP neural network is studied and the fault eigenvector is identified. The simulation example verifies the effectiveness of the method and obtains a high diagnostic rate. (3) considering the complexity of wavelet packet fault feature extraction. In order to further improve the identification degree of fault features, A fault feature vector extraction based on cross wavelet transform is proposed, which makes it unnecessary to do noise reduction and can directly analyze and contrast the amplitude and phase relationship between the two signals at the same time. The original fault feature vector is formed, the final fault vector is obtained by combining the main component analysis, the new fault feature vector is input into the improved BP neural network for fault identification, and the above fault feature extraction methods are analyzed and compared. Simulation examples show that the proposed method is effective and has a higher diagnostic rate.
【学位授予单位】：合肥工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN707

【参考文献】