基于样条逼近和小波分解的Hammerstein模型辨识(英文)

Identification of Hammerstein Model Based on Spline Approximation and Wavelet Decomposition

采用三次B-样条方法取代传统的多项式方法逼近Hammerstein模型中的非线性部分,在保持多项式的简单性和逼近的可行性的同时,大大提高了逼近精度。同时采用小波变换对辨识数据进行多尺度分解,通过在低频空间抑制高频噪声干扰从而实现精确建模。理论分析和仿真结果都表明估计结果具有渐近无偏性和一致收敛性,该方法辨识精度高,具有良好的实用性。

A method is presented to approximate the nonlinear gain of Hammerstein model using the cubic B-spline instead of the traditional polynomial approximation. Compared with the polynomial approximation, not only the method is simple and feasible, but also the accuracy of approximation is improved evidently. To improve identification accuracy, we decompose input and output data by applying the wavelet multiresolution analysis because noises are restrained in low-frequency band. The theory analysis and simulation results show that the estimation is asymptotically unbiased and has strong consistency, and that the new method is very efficient and practical.