一种基于前馈神经网络的 NARMAX 模型辨识新方法

Identification of NARMAX Model Using Feedforward Neural Network

本文提出了一种通过分析非线性动态系统的输入输出进行系统参数辨识的新方法。通过对非线性系统的ＮＡＲＭＡＸ模型和前馈神经网络（ＦＮＮ）模型的比较，得出两者在形式上的等价性，这种等价性使得系统ＮＡＲＭＡＸ模型的参数可以通过一个充分学习的神经网络得到。文中给出了利用隐层采用Ｓｉｇｍｏｉｄ函数的前馈神经网络（ＦＮＮ－Ｓ）辨识ＮＡＲＭＡＲＸ模型参数的完整算法。仿真结果说明，该算法比传统的最小二乘法（ＬＳ）和文献［１］提出的隐层采用多项式函数的ＦＮＮ方法（ＦＮＮ－Ｐ）均具有一定的优越性。

This paper addresses parametric system identification of nonlinear dynamic system by analysis of the input and output signals.The estimation of the system by use of nonlinear autoregressive moving average(NARMAX)model with exogenous inputs is compared with that of the system using feedforward neural network(FNN) model.The equivalence of the FNN model to NARMX model shows that NARMAX coefficients can be obtained from the network provided that the network is properly trained.An algorithm which uses a FNN with sigmoid activation function to estimate NARMAX coefficients is presented in detail.The simulation results show that this algorithm outperforms least squares method and FNN with polynomial activation functions in a previous paper.