本文把具有线性处理单元的神经网络应用于线性时不变动态系统辨识;讨论了用多层前向网络和回归网络分别对线性时不变动态MIMO系统的输入输出模型和状态空间模型进行辨识的方法;提出了用线性系统粗糙模型的先验知识对神经网络权进行初始化的方法.仿真实验表明所提出的初始化方法比随机初始化法的学习有更快的收敛速度. In this paper, the neural network with linear processing unit is applied to the linear time-invariant dynamic system identification. The multi-layer forward network and the regression network are respectively used to identify the input-output model and the state space model of the linear time invariant dynamic MIMO system Method is proposed to initialize the weights of the neural network by using the prior knowledge of the linear system rough model.The simulation results show that the proposed initialization method has a faster convergence rate than the random initialization method.