电力系统保护与动态安全监控教育部重点实验室(华北电力大学)，北京市 昌平区 102206

利用Matlab/Simulink搭建了待辨识励磁系统及单机仿真系统模型，分别采用M序列伪随机信号、机端电压阶跃信号及系统扰动作激励，利用遗传算法辨识励磁系统参数。首先利用多窗口法对不同激励信号进行频谱分析，论证了基于实测扰动的励磁系统参数的可辨识性；然后基于同样扰动，分别以不同采样频率及记录时间(数据长度)进行参数辨识，研究信号采样频率及记录时间对参数辨识结果的影响。结果表明采样频率及记录时间互相配合可以得到较好的辨识精度。

By use of Matlab/Simulink the model of excitation system to be identified as well as that of a single machine simulation system are set up, in which the M serial pseudo-random signal, the step-function signal of generator terminal voltage and system perturbations are taken as incentive signals and the parameters of excitation system is identified by genetic algorithm (GA). Firstly, the multitaper method is used to analyze the power spectrum of different excitation signals to conduct the argumentation of the identifiability of excitation system parameters based on actual measurement of perturbations; then based on the same perturbations, the parameter identifications are respectively performed by various sampling frequencies and record times, namely the data length, to research the impact of sampling frequency and record time on identification results. It is shown that by means of the coordination of sampling frequency with record time a better identification accuracy can be obtained.