摘要
本文提出一种针对非逆稳定随机离散系统的神经网络自校正调节器。为使神经网络调节器有在线自学习能力和更好的适应性与鲁棒性,本文引入衰减激励信号以产生自学习所需的误差信号并激发起系统的各个模态。此外还引入能评价当前控制效果的评价函数以决定是否将在线运行数据选作学习样本来训练神经网络调节器以及决定学习的强度,从而实现了自组织学习与控制,仿真结果表明该方法的有效性。
A neural net self-tuning regulator for non-inverse stable stochastic systems is proposed in this paper. To enhance the on-line self-learning ability, adaptiveness and robustness of the regulation method, an attenuating excitation signal is introduced to produce the error signal for self-learning process and to excit all modes of the systems. Besides, an evaluating function which can evaluate the control efffect is introduced to decide whether the on-line operate data can be chosen as the learning samples to train the neural net regulator and the learning strength, then the self-organized learning and control are realized. Simulation results show the effectiveness of the method.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
1995年第1期64-69,共6页
Pattern Recognition and Artificial Intelligence
关键词
神经网络
非逆稳定系统
衰减激励
自校正调节器
Neural Net, Intelligent Control, Self-Tuning Regulation, Non-Inverse Systems, Attenuating Excitation.
