Nonlinear System Modeling Based on Wavelet Neural Networks

Currently., the modeling of nonlinear systems has garnered widespread research interest. In this paper., a wavelet neural network (WNN) based on Mexican Hat wavelet functions for precise identification of nonlinear system dynamics is proposed. First., a radial basis neural network (RBFNN) with Mexican Hat wavelet functions as the kernel function is designed to enhance the accuracy of nonlinear system modeling. Second., by computing the Euclidean distance between system states and grid points., the grid points were optimized., enhancing the computational performance of the algorithm. Third., to accelerate the convergence speed of the identification error., a switching function for the identification error was defined. Finally., the new algorithm was compared with existing deterministic learning algorithms. Experimental results indicate that the wavelet neural network can achieve more accurate identification of nonlinear dynamic systems.