基于神经网络的NPC三电平逆变器故障诊断方法的研究

发布时间：2018-12-12 21:15

【摘要】：目前,高压、大功率逆变器得到了广泛的应用,其中主要以多电平逆变器为主。虽然多电平逆变器具有器件耐压要求低、等效开关频率高、输出波形好等优点,但也有自身的缺陷。它需要的功率开关器件较多,结构较为复杂,从而运行故障机率增大,不仅增加了逆变器的维护成本,而且降低了系统的工作可靠性。一旦多电平逆变器发生故障会直接影响生产质量和安全,产生严重的损失。因此,多电平逆变器的故障诊断问题,已经成为目前的研究热点,快速诊断出多电平逆变器故障,能有效降低生产损失,具有重要的实际意义和经济价值。本文以多电平逆变器中最常用的NPC三电平逆变器为研究对象,分析了NPC三电平逆变器开路故障后的表现形式,并由此提出了基于小波包变换和神经网络的NPC三电平逆变器故障诊断方法。本文首先从NPC三电平逆变器的拓扑结构和工作原理出发,介绍了NPC三电平逆变器的故障类型,通过Matlab/Simulink软件搭建了故障模型并进行了仿真分析。其次,研究了小波分析和BP神经网络在故障诊断中的应用,采用小波包变换对故障电压信号进行了小波包分解,提取小波包能量值并构造出特征向量,作为神经网络的输入,然后设计了神经网络故障分类器,以此对NPC三电平逆变器进行故障诊断,通过仿真数据验证了本文所提出方法的可行性和有效性。最后,搭建了基于DSP芯片控制的NPC三电平逆变器实验平台,在该平台上对逆变器的各种开路故障进行了实验,并给出了实验故障波形,通过Matlab GUIDE(图形用户界面开发环境)完成了NPC三电平逆变器故障诊断系统的界面设计,实验结果表明该故障诊断系统操作简单、实用性强、具有很高的故障诊断能力,从而也验证了本文所提出故障诊断方法的可行性和有效性。
[Abstract]:At present, high-voltage and high-power inverters are widely used, among which multilevel inverter is the main. Although the multilevel inverter has the advantages of low voltage requirement, high equivalent switching frequency and good output waveform, it also has its own defects. It requires more power switch devices and complex structure, which increases the probability of running failure, which not only increases the maintenance cost of the inverter, but also reduces the reliability of the system. Once the multilevel inverter fails, it will directly affect the quality and safety of production, resulting in serious losses. Therefore, the fault diagnosis of multilevel inverter has become a hot research topic at present. It has important practical significance and economic value to quickly diagnose the fault of multilevel inverter, which can effectively reduce the loss of production. In this paper, the NPC three-level inverter, which is the most commonly used multilevel inverter, is taken as the research object, and the performance of NPC three-level inverter after open circuit fault is analyzed. A fault diagnosis method for NPC three-level inverter based on wavelet packet transform and neural network is proposed. Based on the topology and working principle of NPC three-level inverter, the fault types of NPC three-level inverter are introduced in this paper. The fault model is built by Matlab/Simulink software and the simulation analysis is carried out. Secondly, the application of wavelet analysis and BP neural network in fault diagnosis is studied. Wavelet packet transform is used to decompose the fault voltage signal, extract the energy value of wavelet packet and construct the eigenvector as the input of neural network. Then a neural network fault classifier is designed for fault diagnosis of NPC three-level inverter. The feasibility and effectiveness of the proposed method are verified by simulation data. Finally, the experimental platform of NPC three-level inverter based on DSP chip is built, and various open circuit faults of the inverter are tested on the platform, and the experimental fault waveform is given. The interface design of NPC three-level inverter fault diagnosis system is completed by Matlab GUIDE (graphical user interface development environment. The experimental results show that the fault diagnosis system is simple, practical and has high fault diagnosis ability. Thus, the feasibility and effectiveness of the fault diagnosis method proposed in this paper are also verified.
【学位授予单位】：中国石油大学(华东)
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TM464

【相似文献】