基于网格系的分频段加权正交小波网络辨识非线性系统

WEIGHTED BAND-WISE ORTHOGONAL WAVELET NETWORK BASED ON GRID STRUCTURE FOR NONLINEAR SYSTEMS IDENTIFICATION

在非线性系统辨识中 ,输入数据往往是非均匀分布的 .对于此类问题 ,正交小波网络处理起来较复杂 .本文采用分组中值法对非均匀分布样本进行均匀化处理 ,使正交小波网络的设计和应用能够在网格系上进行 ;然后 ,采用分频段加权技术以便和系统设计相配合 .最后用于辨识非线性动态系统 。

In identification of nonlinear system, the system inputs usually are non-uniformly distributed. The approach using orthogonal wavelet network (OWN)to tackle this problem is complex. This paper transfers non-uniformly distributed data into uniformly-distributed data using grouping median method such that the design and application of orthogonal wavelet network can be completed on the grid structure. In addition, a weighted band-wise OWN is proposed to be combined with system design. The advantages of this method are noise-free, computationally efficient, easy to understand. Finally, a nonlinear dynamic system is identified, the results of simulation show that it is feasible and effective.