论文部分内容阅读
提出了基于最小二乘法原理与FFT相结合的改进型谐波分析方法。基本原理是:提供一组采集信号,利用FFT先计算出谐波的初始参数,然后把初始参数代入模型波形成为其初始参数,计算模型波形和实际波形之间的均方差,如若均方差不满足条件,则进行参数的修正。当均方差最小时,模型波形的参数可以代表实际波形的参数。为避免分析出的谐波次数不准确而出现无效参数,把实际采样数据分成训练组和测试组。在训练组中用最快下降梯度查询学习策略的迭代循环修正参数,在测试组中检测谐波次数的正确性,获得准确的分析结果,实现对次谐波和频率相隔很小的谐波的同步跟踪与分析。通过仿真实验证明了该方法的有效性和准确性。
An improved harmonic analysis method based on least square method and FFT is proposed. The basic principle is: to provide a set of acquisition signals, using FFT to calculate the initial parameters of the harmonic, and then the initial parameters into the model waveform as its initial parameters, calculate the model and the actual waveform waveform of the mean square deviation, if the average variance does not meet Conditions, then the parameters of the amendment. When the mean square error is the smallest, the parameters of the model waveform can represent the parameters of the actual waveform. In order to avoid inaccurate analysis of the harmonic number of invalid parameters appear, the actual sampling data is divided into training and testing groups. In the training group, the quickest descent gradient is used to query the iteration loop of the learning strategy to correct the parameters, and the correctness of the harmonic order is tested in the test group to obtain the accurate analysis result, so that harmonics with minor harmonics Synchronous tracking and analysis. The simulation results show the effectiveness and accuracy of the method.