基于前馈神经网络的电网基波高精度检测  

High Precision Detection of Fundamental of Power Grid Based on Back Propagation Neural Network

作　　者：王勇[1] 付志红[1] 张淮清[1] 王好娜[1] 侯兴哲[2] 

机构地区：[1]输配电装备及系统安全与新技术国家重点实验室(重庆大学),重庆市沙坪坝区400030 [2]国家电网公司电能计量器具性能评估实验室,重庆市渝北区401123

出　　处：《电网技术》2011年第8期124-128,共5页Power System Technology

基　　金：国家自然科学基金项目(40874094)~~

摘　　要：电网基波是电能计量和电能质量评估的重要指标,提出了基于前馈神经网络的电网基波频率和幅值的高精度检测方法。根据数学推导得出:正弦信号过零点与其两侧对称两点的连线与时间轴交点的时间差,同频率满足单调关系,但并非严格的线性关系,而且与幅值无关,据此用前馈神经网络建立该时间差与频率的映射关系。Matlab仿真表明,提出的算法对频率的检测精度达到10?4级,幅值的检测精度高达10?5级,远远高于快速傅里叶变换和Hanning窗的插值算法;随机噪声和谐波对前馈神经网络测量精度的影响很小,该算法具有较强的抗干扰能力。Fundamental of power grid is an important index for electric energy metering and power quality evaluation. A high-precision detection approach, which is based on back propagation neural network （BPNN）, for the frequency and amplitude of power grid fundamental is proposed. It is derived mathematically that the relationship of the time difference, which is between zero-crossing point of sinusoidal signal and the intersection point of time axis and the line connecting two symmetric points on signal curve at both sides of the zero-crossing point, to signal frequency is not strictly linear but monotonous, and the relationship is independent of the amplitude of the signal. Accordingly, the mapping relation between the time difference and fundamental frequency is built by BPNN. Results of simulation by Matlab show that using the proposed algorithm the detection accuracy of fundamental frequency is 10-4 and the detection accuracy of fundamental amplitude is as high as 10-5, and these detection results are sharply higher than those by interpolation algorithms based on fast Fourier transform （FFT） and Hamming window; random noise and harmonics slightly influence the measuring accuracy by BPNN, so the proposed algorithm possesses strong anti-interference capability.

关 键 词：电网基波 前馈神经网络 基波频率 基波幅值 

分 类 号：TM714[电气工程—电力系统及自动化]