节点文献
基于神经网络的逆变器的研究
Research on Inverter Based on Neural Network
【作者】 元志;
【导师】 王云亮;
【作者基本信息】 天津理工大学 , 控制理论与控制工程, 2008, 硕士
【摘要】 逆变器在当今社会具有很广泛的应用,随着电力电子技术的快速发展,大量高功率开关器件相继出现,为逆变器在各行各业的应用提供了保障。本文研究的逆变器主电路采用三相桥式结构。在控制策略方面,传统的正弦脉宽调制(SPWM)具有直流电压利用率较低的缺点,空间矢量脉宽调制(SVPWM)是一种优化的脉宽调制技术,能明显减小逆变器输出电流的谐波成分并能提高电压的利用率,已有取代传统SPWM的趋势,本文即采用电压空间矢量脉宽调制作为控制策略。本文首先对电力电子技术及逆变器的相关知识作了简要的介绍。对空间电压矢量脉宽调制的工作原理做了深入的研究,并且在此基础上构建了基于SVPWM算法的逆变器仿真模型,得到了仿真结果。之后本文提出了基于神经网络的逆变器的构想,在计算逆变器三相导通时间时,运用了3层前馈网络,并采用Levenberg-Marquardt算法对神经网络进行训练。最后,在MATLAB/Simulink环境下,对系统进行了仿真研究,仿真结果表明基于神经网络的SVPWM控制方法是可行的,能达到预期的效果,同时通过合理的运用神经网络,可以提高PWM的频率,减小输出电压中的谐波含量。
【Abstract】 The inverter is used widely in society. With the rapid development of power electronics technology, a large amount of high-frequency switching equipment are appeared which provide the guarantee for the inverter’s application in all kinds of fields. The inverter’s main circuit discussed in this paper adopts the three-phase bridge structure. In terms of controller strategy, the traditional method of SPWM has the setback which is lower electric source utilization ratio. SVPWM is a kind of optimized method, it can reduce the output current’s harmonic component and improve the electric source utilization ratio, having the trend of replacing the traditional SPWM method, this paper just adopts the SVPWM as the controller strategy.This paper firstly introduces the power electronics technology and inverter’s corresponding knowledge, and the paper deeply researches the work’s principle of the SVPWM and establishes the inverter’s simulation model based on the arithmetic of SVPWM, achieved the simulation result. And then this paper puts forward the inverter’s imagination based on neural network, when calculating the three-phase times, the paper uses a three-layer forward-feed network, adoptes the arithmetic of Levenberg-Marquarde to train the network. At last, in the environment of MATLAB/Simulink, the paper makes the simulation research to the system, the simulation results shows that the method of SVPWM based on neural network is feasible and it can achieve the anticipated effect, at the same time by using the neural network rationally, the method can enhance the PWM frequency and reduce the harmonic of output voltage.
【Key words】 Inverter; Space Vector; Neural Network; MATLAB/Simulink;