微网多逆变器并联的故障诊断研究

发布时间：2018-07-09 10:02

本文选题：微网 + 多逆变器并联　；参考：《南昌大学》2017年硕士论文

【摘要】：微网是目前分布式发电发展的自然延伸,也是分布式发电最新最有效的应用形式。随着微网的大量应用,人们对微网系统的可维护性提出了很高的要求。在微网中,逆变器故障的问题是越来越突出,而因其故障带来的危害和影响却是巨大的,因此研究微网环境下的逆变器故障检测与诊断意义重大。论文在深入研究分析近几年最常见的逆变器的故障特征提取与诊断算法的基础上,详细分析了单台逆变器和微网系统中的逆变器断路的故障特点及其故障表现,提出了微网平台下的故障诊断机制;在此基础上,应用小波变换和稀疏编码算法在故障特征提取环节中,最后基于BP神经网络分类器完成对微网系统中逆变器断路故障的诊断。论文首先从微网系统的结构及工作原理入手,在MATLAB/Simulink下搭建了微网系统的仿真模型,并验证了平台的正确性,为后续的故障分析及诊断工作奠定了基础。接着在对微网多逆变器并联系统下的逆变器故障分析之前,对单逆变器的故障特点及表现进行分析并给出了典型故障的仿真波形,得出故障表现与故障类型和故障发生的时间有关系;据此对微网下的逆变器进一步分析,得出在微网系统某逆变器故障时各逆变器模块在微网并网模式下的独立性及离网模式下的相关性,说明了故障诊断的复杂性。所以本文提出微网中每台逆变器应具备故障自检的机制,并且这种机制应该是一种可以兼容两种模式并具备排他故障能力故障诊断机制。为获取故障表现的本质特征,需要采用故障特征提取算法,本文分别采用了小波变换和稀疏编码算法对微网下的逆变器故障进行特征提取,得到各自的特征向量,从而获得各自的总样本集。然后分别使用相同编号组成的训练样本集训练BP神经网络,得到2个训练好的BP神经网络分类器,并用剩下相同编号组成的测试样本集进行故障诊断,结果表明,基于这两种算法诊断结果都具有较高的准确率,验证了故障诊断机制的可行性和正确性;而稀疏编码得到的故障特征向量却表现出更加高的区分度,诊断结果的准确率也更高,表明在微网下的逆变器故障诊断中,稀疏编码算法更加具有合理性和可取性。
[Abstract]:Micro network is the natural extension of the development of distributed generation and the latest and most effective application form of distributed generation. With the application of the micronetwork, the maintainability of the micronetwork system is very high. In the micro network, the problem of the inverter fault is becoming more and more prominent, but the harm and influence caused by its failure are huge. Therefore, it is of great significance to study the fault detection and diagnosis of the inverters in the micronetwork environment. Based on the in-depth study and analysis of the most common fault feature extraction and diagnosis algorithms of the inverters in recent years, the fault characteristics and fault manifestations of the inverter in the single inverter and Microgrid are analyzed in detail. The fault diagnosis mechanism under the network platform; on this basis, the wavelet transform and the sparse coding algorithm are used in the fault feature extraction. Finally, based on the BP neural network classifier, the fault diagnosis of the inverter in the micro network system is completed. First, the paper starts with the structure and working principle of the micronetwork system, and builds it under the MATLAB/Simulink. The simulation model of the microgrid system is used to verify the correctness of the platform, which lays the foundation for the subsequent fault analysis and diagnosis. Then, the fault characteristics and performance of the single inverter are analyzed and the typical fault simulation waveform is given before the fault analysis of the inverter parallel system. The fault performance is obtained. It is related to the type of fault and the time of the fault. According to the further analysis of the inverters under the micro network, the independence of the inverter modules in the microgrid connection mode and the correlation in the off network mode are obtained when some inverters in the micronet system fail, and the complexity of the fault diagnosis is explained. So this paper proposes each inverter in the micronet. There should be a mechanism for self inspection of fault, and this mechanism should be a fault diagnosis mechanism that can be compatible with two modes and has the ability of troubleshooting. In order to obtain the essential features of the fault performance, the fault feature extraction algorithm should be adopted. The wavelet transform and sparse coding algorithm are used to carry out the inverter fault in the micronetwork respectively. The feature vectors are extracted and the respective feature vectors are obtained, and then the total sample sets are obtained. Then the training samples set up with the same number are used to train the BP neural network, and 2 trained BP neural network classifiers are obtained, and the test samples set up with the same number are used to diagnose the fault. The results show that the two algorithms are based on these algorithms. The diagnosis results all have high accuracy, which verifies the feasibility and correctness of the fault diagnosis mechanism, while the fault feature vector obtained by sparse coding shows a higher degree of segmentation and the accuracy of the diagnosis results is higher. It shows that the sparse coding algorithm is more reasonable and advisable in the fault diagnosis of the inverters under the micronet.
【学位授予单位】：南昌大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM464

【参考文献】