输出过采样闭环系统辨识方法中最优过采样率的选择

提出了一种基于输出过采样技术的线性离散时间闭环系统辨识方法 ,通过对输出端施加过采样 ,将原闭环系统模型转化为过采样模型 .对于输出端白噪声、有色噪声干扰 ,分别利用最小二乘辨识算法和误差预报算法辨识出过采样模型 ,进而计算出原闭环系统的模型参数 .辨识结果中模型参数的估计误差服从均值为0的正态分布 ,方差由过采样率决定 ,据此推导出了使参数估计误差最小化的最优过采样率的计算方法 .该方法实现简单 ,运算量小 ,估计精度高 .仿真实验表明 ,当信噪比大于 15dB时 ,该方法的估计精度可达98% ;当过采样率为 15～ 2 0时 ,算法具有最优的辨识精度

Selection of Optimal Over-Sampling Rate in Close-Loop Identification with Output Over-Sampling

A new identification method for a linear discrete-time noisy closed-loop system was proposed based on the output over-sampling technique. According to the relationship between the selection of over-sampling rate and the estimation accuracy, the calculation method of optimal over-sampling rate was obtained. When the system output was over-sampled, the original system model will be changed to a new over-sampled model. For the output with white noise and colored noise, a general least squares method and prediction error method are applied respectively to identify the over-sampled model. By the connection of the over-sampled model and the original model, the original close-loop model can be recognized. The distribution of the estimation accuracy under different over-sampling rate was studied and the calculation method of the optimal over-sampling rate was proposed. Compared with conventional algorithms, the method is easy to be realized, with less operating load and high estimation accuracy. Numerical simulation shows that the estimation accuracy is 98% when the signal-to-noise ratio is bigger than 15 dB and the optimal estimation accuracy is obtained while the over-sampling rate is between 15 and 20.