滤波器-神经网络的谐波检测方法

An approach with filters-neural network for harmonic detection

电网谐波的高精度检测是电能计量和电能质量评估的基础。针对神经网络的谐波检测算法中,计算精度受基波频率精度影响较大的问题,提出用数字滤波结合牛顿反插值算法得到高精度的基波频率,然后用线性神经网络算法检测电力系统各次谐波的频率、幅值和相位。计算结果表明,该算法在频率波动和白噪声干扰的情况下,依然能得到高精度的谐波参数信息,其精度远高于FFT算法与加汉宁窗的FFT算法,在电力系统谐波测量中有一定的应用价值。

High-precision detection of harmonic is the basis of the assessment of energy metering and power quality.Aiming at the problem that the accuracy of algorithms of neural network harmonic detecting is largely affected by the fundamental frequency accuracy,this paper presents the digital filter combining with Newton＇s inverse interpolation algorithm to get fundamental frequency with high accuracy,then uses linear neural network algorithm to detect frequency,amplitude and phase of power system harmonics.The results show that under the interference of frequency fluctuation and white noise,it can still get accurate harmonic parameters,whose accuracy is much higher than that of the FFT algorithm and FFT algorithm with the Hanning window,and it has certain application value in power system.