基于改进的ARMA递推算法的低频振荡模式在线辨识  

Improved ARMA recursive algorithm for online identification of low frequency oscillation modes

作　　者：陈刚 龚啸[2] 李军[3] 古志明[3] 

机构地区：[1]南方电网科学研究院,广东广州510080 [2]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400044 [3]重庆市电力公司,重庆400015

出　　处：《电力系统保护与控制》2012年第1期12-17,49,共7页Power System Protection and Control

摘　　要：为提高电力系统低频振荡现象的实时监测水平,提出采用一种基于自回归滑动平均模型的两段加权递推最小二乘算法进行低频振荡模式辨识,并通过估计ARMA谱的方法以提取低频振荡的主导模式。该改进算法先采用加权递推最小二乘算法拟合高阶AR模型单独得到白噪声估值,再将该估值用于常规加权递推最小二乘算法中,提高了算法参数辨识的精度和收敛速度。New-England 39节点系统的时域仿真测试验证了该改进算法对低频振荡模式辨识的有效性,并通过与常规加权递推最小二乘算法辨识效果的比较验证了该改进算法对低频振荡模式的辨识具有更好的精确性且提高了收敛速度。最后通过对某电网PMU实测数据的辨识分析,验证了该改进算法能够准确地辨识系统的低频振荡主导模式频率和阻尼比,具有实际的工程意义。In order to improve the level of real-time monitoring of low frequency oscillation, this paper proposes to use a two parts weighted recursive least square （WRLS） algorithm based on auto-regressive moving-average （ARMA） model to estimate the low frequency oscillation modes, and extracts the domain modes of low frequency oscillation by the method of ARMA spectrum estimation. The improved algorithm uses the obtained white noise estimates by fitting the higher order autoregressive （AR） model by the WRLS method in the conventional WRLS method, and then it has the preferable accuracy and the fast convergence rate of parameter identification. The validity of proposed algorithm is demonstrated with simulation data from New-England 39-bus system. Comparison with proposed algorithm and conventional weighted recursive least square algorithm shows the advantages of the algorithm in this paper. Finally, identification analysis of practical signal measured of PMU in some grid demonstrates that the proposed algorithm can accurately estimate the frequency and damping ratio of power system low frequency oscillation domain modes, so the proposed algorithm has the practical project significance.

关 键 词：自回归滑动平均模型 加权递推最小二乘算法 ARMA谱 低频振荡在线辨识 主导模式 

分 类 号：TM712[电气工程—电力系统及自动化]