摘要
针对电力系统低频振荡模式识别问题,本文提出了一种用于电力系统低频振荡模式识别的指数阻尼正弦神经网络(EDSNN)方法。在采用快速傅立叶变换进行排序后,通过引入了拓扑结构与低频振荡(LFO)信号的数学表达式完全一致的指数衰减正弦模型,将参数估计转化为优化问题。仿真结果表明,通过对数值信号、模拟电力系统信号和实际测量信号的应用,该方法在抗噪声能力、参数精度和计算速度等方面均优于现有的自适应线性神经元(Adaline)模式识别方法。
Aiming at the problem of pattern recognition of low frequency oscillation in power system,an exponential damped sinusoidal neural network(EDSNN)method is proposed for pattern recognition of low frequency oscillation in power system.After sorting by fast Fourier transform,the exponential attenuation sinusoidal model,whose topological structure is identical with the mathematical expression of LFO signal,is introduced to transform parameter estimation into optimization problem.The simulation results show that the method is superior to the existing adaptive linear neuron(Adaline)pattern recognition methods in terms of anti⁃noise ability,parameter accuracy and computing speed by applying numerical signals,analog power system signals and actual measured signals.
作者
孙福寿
SUN Fu-shou(State Grid Jilin Electric Power Co.,Ltd.,Changchun 130000,China)
出处
《电子设计工程》
2020年第12期120-124,129,共6页
Electronic Design Engineering
关键词
模式识别
低频振荡
神经网络
指数阻尼正弦
pattern recognition
low frequency oscillation
neural network
exponentially damped sine