时滞系统神经网络逆控制研究

Neural Network Inverse Control in Delay System

为了改善大时滞对象的控制效果,提出一种带时滞时间辨识的神经网络逆控制系统。利用自适应线性元件与BP网络相结合,辨识对象的时滞时间及不合时滞环节的模型,再对不合时滞的模型构造神经网络逆,并选择合适的参考模型使逆模型的输出平滑。将训练好的逆模型作为控制器,与被控对象串联形成开环控制,有效避免了闭环控制可能引起的不稳定。仿真结果表明,该控制策略能够实现系统快速平稳的输出,且能够克服时滞时间及参数变化引起的不良影响,与Smith预估控制器相比,具有较好的鲁棒性及抗干扰能力。

In order to improve the control effect of the large time-delay system, a neural network inverse control system with delay time identification is proposed. By using adaptive linear element combined with BP neural network, the delay time and the positive model without delay are indentified, its neural network inverse model are obtained, and the appropriate reference model is selected to smooth the output of the inverse model. The trained inverse model is used as a controller and with controlled object in series to form open-loop control, so the instability caused by the closed-loop control systems is avoided. Simulation results show that the neural net- work inverse control with delay time identification can get rapid and smooth output, and be able to overcome the adverse effects caused by the time delay and the parameters variations. Compared with the Smith predictive controller, its robustness and disturbance rejection performance are improved.