基于RBF神经网络的自回归系统辨识模型

根据神经网络能以任意精度逼近任意非线性连续函数的特点,通过径向基函数神经网络构建非线性动态系统的辨识模型。针对该模型输入值超出径向基函数的映射区域时将导致系统辨识输出值为零的现象,提出了一种基于改进径向基函数结构的自回归系统辨识的方法,有效地消除了零现象。这使得自适应辨识模型在较大的输入向量下能够逼近实际系统的输出,从而提高了系统辨识的鲁棒性。该方法的可行性得到了仿真验证。

Identification Model of Regression System based on RBF Neural Network

Nonlinear dynamic system identification model with radial basis function neural network is developed based on the function that neural network can close to nonlinear continuous function with high precision.The method of autoregressive system identification based on improved radial basis function structure is proposed with respect to zero phenomenon with input larger than mapping area of radial basis function,which can address this problem with good performance.The adaptive identification model can close to output of system even though a large vector inputted.In addition,the robustness of system identification is improved.This feasibility of method is validated through simulation.