一种非线性系统的鲁棒型预测控制算法

提出一种基于神经网络的鲁棒型广义预测控制(GPC)方法,该方法首先用神经网络对非线性系统进行辨识,然后在控制中将模型输出值与测量输出值进行综合,代替量测输出用于控制中,从而降低辨识器与控制器对未建模动态的敏感性,增强控制器的适应能力和鲁棒性.仿真结果表明:将本方法应用于非线性系统控制,对未建模动态具有较强的鲁棒性和控制能力.

A robust predictive control method for nonlinear system

A robust Generalized Predictive Control(GPC) based on neural network model is presented in the paper.Firstly the nonlinear system by neural network is identified.Then the output sequence used in the controller design is replaced by the combination of the system output and the model output,so the sensitivity of the identifier and controller can be cut down,and the adaptive ability and robustness of the controller are increased.The simulations show that it suits to the nonlinear system,and has strong robustness and great control ability of the unmodeling dynamic.