时变大纯滞后系统的单神经元自适应控制

Single-neuron Adaptive Control for Time-variable Large Delay Systems

阐述一种新型单神经元自适应控制器,对时变大纯滞后系统实现快速有效的实时控制。该单神经元采用一种新学习算法,并与Smith补偿、在线辨识相结合,在保留单神经元器适应性强优点的同时,改善了单神经元器的动态性能,减轻了大滞后对象控制结果不能及时反馈的不足。应用该控制策略对大滞后一阶仿真研究表明,对大滞后时变系统具有较强的适应性和鲁棒性,各种控制性能优于常规单神经元PID和常规PID。

This paper presents a new kind of Single-neuron PI controller for time-variable large delay systems.The controller is made up of a single-neuron and Smith predictor.The model of Smith predictor is indentifie online.Especially a new weight-learning algorithm is adopt for the single-neuron,which greatly improved the learning speed of the single-neuron comparing to Hebb algorithm.Experiments with single order delay system verified the controller processes self-turn and robustness,and it's overall control performance is better the PI controller and single-neural predictive controller with Hebb algorithm.