基于MSWF和改进Adaline神经网络的间谐波分析  

Interharmonic analysis based on multi-stage wiener filter and improved Adaline neutral network

作　　者：陈国志[1] 陈隆道[1] 蔡忠法[1] 

机构地区：[1]浙江大学电气工程学院,浙江杭州310027

出　　处：《电力自动化设备》2010年第4期55-58,共4页Electric Power Automation Equipment

基　　金：浙江省教育厅科研项目(Y200803502)~~

摘　　要：为降低计算复杂度,将多级维纳滤波器(MSWF)应用于电力系统间谐波分析,提出基于MSWF和改进自适应神经网络的间谐波分析方法。利用MSWF的前向递推实现信号子空间和噪声子空间的快速估计,不需要估计数据的协方差矩阵及其特征值分解,减小了计算量。应用新的最小描述长度准则和TLS-ESPRIT算法确定谐波、间谐波的个数及频率。为提高收敛速度,应用基于递归最小二乘学习算法的自适应神经网络分析谐波和间谐波的幅值和相位。Matlab仿真结果验证了所提算法的有效性。该方法计算复杂度低,分辨率高,精度高,收敛快。An interharmonic analysis algorithm based on MSWF（Multi-Stage Wiener Filter） and improved Adaline neural network is proposed to reduce the computational complexity for the harmonic and interharmonic analysis of power system.The signal subspace and noise subspace are quickly evaluated by the forward recursion of MSWF,which avoids the covariance matrix and its eigenvalue decomposition to significantly re-duce the computational complexity.The number and frequencies of harmonic and interharmonic are deter-mined by the novel minimal description length principle and TLS-ESPRIT algorithm.The amplitude and phase of harmonic and interharmonic are estimated by the improved Adaline neutral network based on re-cursive least square to speed up the convergence.The results of Matlab-based simulation prove the proposed algorithm is of low computational complexity,high resolution,superb precision and fast convergence.

关 键 词：间谐波 多级维纳滤波器 神经网络 最小描述长度准则 递归最小二乘法 

分 类 号：TM714[电气工程—电力系统及自动化]