基于神经网络的三电平逆变器故障诊断研究

发布时间：2018-03-08 11:55

本文选题：三电平逆变器　切入点：故障诊断　出处：《安徽工程大学》2015年硕士论文　论文类型：学位论文

【摘要】：多电平逆变器具有克服传统逆变器开关应力大、直流母线电压利用率低等优点,尤其适用于高压、大功率场合的应用。本文首先介绍多电平逆变器的产生背景、突出优点和硬件拓扑结构,并总结了其控制策略的发展历程。然后通过介绍基于信号处理、神经网络、支持向量机这三种电力电子装置主要的故障诊断方法,指明多电平逆变器故障诊断技术目前很不成熟,并揭示研究难点主要集中在多电平逆变器的建模和制造成本、中点电位波动、死区补偿和窄脉冲、无源软开关这四方面。指明解决多电平逆变器故障诊断这一历史性难题要把重点放在拓扑结构和控制策略等机理上。通过详细阐述多电平逆变器技术未来的五大发展趋势,表明多电平逆变器具有广泛的应用前景。三电平逆变器作为多电平逆变器的一种,在高压大功率场合获得了较为广泛的应用。但是功率器件的增加,提高了系统的故障率,因此,如何保证系统安全可靠运行便成为了一个重点研究的问题。本文以二极管钳位式三电平逆变器为例,分析了三电平逆变器在功率管开路下的电路工作情况及故障表现形式,并由此提出了一种基于神经网络的三电平逆变器功率器件开路故障的检测和诊断方法。本文首先从三电平逆变器的拓扑结构和工作原理出发,分析了三电平逆变器主电路的故障类型,将主电路功率开关器件的开路故障分为九大类,应用MATLAB软件对这些故障进行了仿真分析,并建立了简单故障诊断表。其次,采用傅立叶变换的方法提取各种复杂故障波形的各次谐波分量作为故障诊断的特征参数,然后对故障波形的频谱特征进行分析,将其用于故障样本的编码中,从而建立了基于神经网络的故障诊断模型。将神经网络的诊断方法与小波分析的诊断方法结合起来进行故障诊断。最后,设计了基于DSP的三电平逆变器的故障诊断实验平台,构建了三电平逆变器的主电路,设计了DSP的外围电路,包括信号缓冲保护电路、驱动电路、电压电流检测电路、辅助电源等。编写了三电平逆变器的SVPWM信号发生程序和故障诊断程序等。实验结果表明,本文提出的基于神经网络的三电平逆变器故障诊断方法具有一定的可行性和实际应用价值。
[Abstract]:Multilevel inverter has the advantages of overcoming the high switching stress of conventional inverter and low utilization ratio of DC bus voltage. It is especially suitable for high voltage and high power applications. The background of multilevel inverter is introduced in this paper. The main fault diagnosis methods based on signal processing, neural network and support vector machine are introduced. It is pointed out that the fault diagnosis technology of multilevel inverter is not mature at present, and it is revealed that the research difficulties mainly focus on the modeling and manufacturing cost of multilevel inverter, neutral point potential fluctuation, dead-time compensation and narrow pulse. The four aspects of passive soft switching. It is pointed out that in order to solve the historic problem of multilevel inverter fault diagnosis, the emphasis should be placed on topology structure and control strategy, and the five development trends of multilevel inverter technology in the future are described in detail. As a kind of multilevel inverter, three-level inverter has been widely used in high-voltage and high-power applications. However, with the increase of power devices, the failure rate of the system is improved. Therefore, how to ensure the safe and reliable operation of the system has become a key issue. This paper takes diode clamped three-level inverter as an example. In this paper, the circuit operation and fault performance of three-level inverter under open circuit of power transistor are analyzed. Based on neural network, an open circuit fault detection and diagnosis method for three-level inverter is proposed. Firstly, the topology and working principle of three-level inverter are discussed. The fault types of main circuit of three-level inverter are analyzed. The open circuit faults of power switch devices of main circuit are divided into nine categories. These faults are simulated and analyzed by MATLAB software, and a simple fault diagnosis table is established. The Fourier transform method is used to extract the harmonic components of various complex fault waveforms as the characteristic parameters of fault diagnosis, and then the spectrum features of the fault waveforms are analyzed and used in the coding of fault samples. The fault diagnosis model based on neural network is established. The fault diagnosis method of neural network is combined with the diagnosis method of wavelet analysis. Finally, the fault diagnosis experiment platform of three-level inverter based on DSP is designed. The main circuit of three-level inverter is constructed, and the peripheral circuit of DSP is designed, including signal buffer protection circuit, driving circuit, voltage and current detecting circuit. The SVPWM signal generating program and fault diagnosis program of three-level inverter are programmed. The experimental results show that, The fault diagnosis method based on neural network for three-level inverter has certain feasibility and practical application value.
【学位授予单位】：安徽工程大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TM464

【参考文献】