基于人工神经网络对谐波鉴幅

电子机电论文编号:JD565 附开题报告,论文字数:16155,页数:40

目   次

1  引言 …………………………………………………………………………………1
1.1  电网谐波鉴幅的意义 ………………………………………………………2
    1.2   现有鉴幅方法的优点和不足  ……………………………………………3
    1.3   目前应用研究概况  ………………………………………………………4
2  电网谐波产生的原因、危害、及影响   …………………………………………6
    2.1    谐波的种类及产生的原因   ……………………………………………6
    2.2    谐波的危害及影响    ………………………………………………… 6
    2.3    研究的目标和内容   ……………………………………………………7
3 对工具函数的介绍  …………………………………………………………………9
3.1   对MATLAB工具的介绍 ……………………………………………………9
3.2   对神经网络的介绍 ………………………………………………………11
利用神经网络对电力谐波幅值的测量方法………………………………………14
4.1   简单介绍 …………………………………………………………………14
4.2  神经元结构  ………………………………………………………………14
4.3  感知器的结构及学习算法  ………………………………………………14
4.4   采样频率的要求 …………………………………………………………16
4.5   电力谐波测量的ANN方法的原理 ………………………………………16
4.6   基于多层前馈神经网络的谐波测量电路 ………………………………17
4.7   训练样本的形成方法 ……………………………………………………19
4.8   仿真训练及结果 …………………………………………………………20
4.9   仿真研究 …………………………………………………………………21
结论 ………………………………………………………………………………… 25
致谢 ………………………………………………………………………………… 27
参考文献………………………………………………………………………………28
表1……………………………………………………………………………………29
  附录  ……………………………………………………………………………31

摘要
人工神经网络用于谐波、基波等各种波形的识别是一个新的研究领域。针对神经网络的这一新方法引入到电网谐波幅值测量中，并就神经网络如何构建问题进行了较为系统的研究。
针对电网谐波监测系统，论述了相关人工神经网络的模型构造和学习算法。根据这一课题通过构造一个特殊的多层前向神经网络，根据并行测量的原理，设计了相应的训练过程。通过对BP算法及其改进型算法的研究，将该算法应用于3次、5次谐波的幅值测量中。针对电网谐波监测系统给出了网络的训练方法和步骤，研究了训练样本的形成方法，并对训练样本的输入方式进行了研究。最后对训练好的神经网络进行了仿真，仿真结果表明针对这个课题所提出的基于人工神经网络监测方法的有效性和实用性。

关键词 人工神经网络，测量，谐波电流

Title   According to the artificial neural network examines to the harmonic value                                                           

Abstract
The artificial neural network used for fundament and harmonic which it is a new research realm. This text neural network of this a the new method leads into to the charged barbed wire net harmonic value to measure, combining how the neural network set up the problem proceeded than for the research of the system.
This text is a model to aim at charged barbed wire net harmonic monitors system, discussed the related artificial neural network structure primarily with study the calculate way. This text passes before constructing a special and many layers face the neural network, is measured according to the principle that proceed together, designed the homologous training process. Pass to the research between the calculate way of BP and its improvement type calculate way, apply this calculate way in 3 times and 5 times value of harmonic measures the inside. This text gives a training for network method with the step, studied the formation method of the training sample, and proceed the research to the importation method that trained the sample. Imitate to the nerve network that train the good proceeding finally imitate true, imitate true the result expresses this text puts forward of according to the artificial neural network monitors the usefulness of the method with function.

Keywords  Artificial neural network, measure, harmonic current