稳定平台神经网络扰动补偿内模控制

IMC and ESO based on ANN of stabilized platform

为了避免传统内模控制器和扰动观测器中参数选取难的问题，同时适应稳定平台非线性的特性要求，采用小波神经网络对内模控制器的滤波时间常数、扰动观测器的补偿系数进行在线寻优。应用小波神经网络对非线性函数的任意逼近特性，对稳定平台的未建模动态和总和扰动进行有效估计。对提出的小波神经网络扰动观测器内模控制器进行仿真研究，相比传统控制器，采用小波神经网络扰动观测内模控制器的稳定平台，系统在1．5Hz以内扰动隔离能力提升一倍以上。

In order to avoid the difficulty of parameter selection in the traditional internal model controller and disturbance observer, at the same time to adapt the nonlinear characteristics of stabilized platform, this paper adopts the wavelet neural network internal model controller compensate the coefficient of disturbance observer and filter time constant of internal model controller. By using the arbitrary nonlinear function approximation properties of wavelet neural network, we can estimate the unmodeled dynamics and combined disturbance effectively. Simulation studies show that the control performance of the Internal model controller by using wavelet neural network disturbance observation are better than those of the conventional PID controller.