基于FFT和神经网络的高精度谐波分析

A high accurate harmonic analysis method based on FFT and neural network in power system

为了精确分析整数和非整数次谐波,提出了基于快速傅里叶变换(FFT)和神经网络的谐波分析方法,该方法的特点是采用基函数参数可调的神经网络。具体是先把信号进行FFT处理,得到谐波个数和精度不高的谐波幅值、相位、谐波次数;其次根据谐波个数设定神经元的个数,根据预处理后得到的幅值、和相位、谐波次数设定神经网络权值和基函数参数迭代的初始值;最后对人工神经网络进行训练,便可实现整数和非整数次谐波的精确分析,同时能将频率相近的非整数次谐波分离。仿真结果验证了该方法的有效性与易实现性。

A high accurate algorithm is presented for analysis of integer harmonic and non-integer harmonic in power system, which is based on FFT and neural network. A characteristic of the method is that adjustable parameter base function is adopted by neural network. First, the sampled signal is processed with FFT algorithm.By this algorithm, the number, magnitudes, phases, and orders of harmonics are obtained. Second, applying results analyzed with FFT, the number of neural nodes is established according to the number of harmonics, the initial weights of neural network are magnitudes of harmonics, the iterative initial parameters of base function are phases and orders of harmonics. Finally, by training artificial neural network, integer harmonics and non-integer harmonics can be analyzed precisely, at the same time, the close non-integer harmonics can be separated. The simulation results verify the effectiveness and practicability of the algorithm.