基于连续时间方法的励磁系统参数辨识

IDENTIFICATION OF EXCITATION SYSTEM PARAMETERS BASED ON CONTINUOUS-TIME APPROACHES

详细分析了用于励磁系统参数辨识的连续时间辨识方法～一分段线性多项式函数(PLPF)法,指出PLPF法本质上是在整个时间区间内采用复化梯形公式的数值积分方法,并讨论了PLPF法的初始条件影响问题,在此基础上,引入了另一种连续时间辨识方法,线性积分滤波(LIF)法,使用UF法可以完全消除积分方程初始条件的影响,即在非零初始条件下无需辨识初始条件依然能得到准确的结果。仿真实验中,就PLPF法及LIF法在初始条件问题和测量噪声影响方面进行了比较讨论,结果表明LIF法与PLPF法相比有一定的优越性。

This paper analyzes in detail a continuous-time identification approach, piecewise linear polynomial functions (PLPF) approach which has been proposed for parameter estimation of excitation system and proved to be essentially a numerical integration approach wherein the trapezoidal integrating rule is used over the whole time interval. The initial condition problem introduced by the multiple integrations in PLPF method is also discussed. At the same time, a new continuous-time identification approach, the linear integral filter(LIF) approach is presented which can be used to eliminate completely the effect of the initial condition problems of the integral-equation, that is to say, it does not require to estimate the initial conditions under the non-zero initial conditions. In a simulation study, both PLPF approach and LIF approach are compared based on experiments in which the influence of initial condition problem and measurement noises are involved. The results show that the LIF method is superior to PLPF method in a sense.