摘要
对于参数变化且含有时滞的被控对象,传统控制方法难以获得理想效果.本文基于神经元模型及学习策略,提出采用不同学习速率的神经元学习方法,由此构成的神经元控制器能加速神经元权值收敛,改善神经元控制品质,使这类被控对象可取得满意的控制效果.仿真结果表明,只要选择适当的神经元参数,可以较明显地减少神经元控制的超调量,并有很强的鲁棒性和抗干扰性.
For the controlled objects with varying parameters and time lag characteristic ideal effects can not be obtained by traditional control method. Based on neuron model and learning strategy the paper states that satisfactory control result can be accomplished to such objects by the neuron controller which is adopting neuron learning method with different learning rates. The controller makes the convergence of the weight of the neuron faster and improves the control quality. Thus satisfactory control effects are obtained for such controlled objects. The simulation results show that the control overshoot can be decreased obviously by selecting proper neuron parameters. Meanwhile high robustness and anti - interference capability are also featured.
出处
《自动化仪表》
CAS
北大核心
1998年第12期4-9,共6页
Process Automation Instrumentation
基金
国家自然科学基金资助项目(No.69774023)
关键词
神经元控制器
学习速率
时滞特性
仿真
Neuron controller Learning rate Time lag characteristic Simulation
