基于最优线性滤波的PMU测量Ⅲ:M类算法设计

PMU Measurement Based on Optimal Filtering Ⅲ: M Class Filter Design

本系列论文的前2篇,分别研究同步相量测量单元测量的最优滤波原理和P类测量最优算法的具体设计。该文在此基础上,研究M类测量最优算法的具体设计。与P类测量相比,M类测量是在更高准确度基础上兼顾快速响应,并且增加了公认为最难实现的对带外干扰的抑制要求。为了实现从基波通带到带外干扰阻带的快速衰减,基于奈奎斯特频域抽样定理建立动态相量线性参数模型,并采用以凯撒窗为权重的加权最小二乘算法进行模型参数估计,由此构造原型滤波器,其中频域抽样点个数和凯撒窗阶数构成自由参数集。针对数据上报率分别为50、25和10Hz三种情况,求解最优化问题,给出不同数据长度下最优解的分布,并结合快速响应性能选取最优滤波器长度,构成完整的M类测量优化算法。通过仿真试验评估优化算法的实际测量性能,并与IEEE测量标准提供的参考算法结果进行比较,验证所提出优化算法在测量准确度、使用数据长度和单相测量方面具有的综合性能优势。

In Part I and Part II of the three-part paper set, optimal filter design criteria for phasor measurement unit measurement, and practical method designing optimal algorithm for P class have proposed respectively, Part III continues to deal with the practical design of optimal algorithm for M-class measurement. As compared with P class, M class is required to be of higher precision, and an additional ability of rejecting out-of-band （OOB） interference, where OOB rejection is generally acknowledged as most challenging. In order to achieve fast decay from the edge of fundamental passband to the OOB stopband, the prototype filter was constructed from the dynamic phasor model based on frequency domain Nyquist sampling theorem and the parameter estimator based on Kaiser window weighted least square （WLS） method. The free parameter set to be optimized consists of the dynamic phasor model order and the Kaiser window order. The optimization results as functions of data length were provided for the cases of reporting rate equals to 50, 25 and 10Hz respectively. The optimal M-class algorithm for each reporting rate was recommended trading off measurement precision and data length. Excellent performance of the proposed optimal algorithms were verified with simulations, and their superiority in measurement precision, response time, and single-phase applicability is highlighted as compared with the reference algorithms recommended in IEEE.C37.118.1 a.