摘要
在用神经网络对PID控制器的控制参数进行优化时,网络的学习速率η和动量参数α的调整尚没有成熟的方法,针对这一问题提出了一种实时调整η和α的模糊归一化算法,把综合误差变化量进行归一化、模糊量化处理之后,η和α根据模糊子集中的隶属函数找出其对应的隶属度。将这一算法应用于非线性系统的解耦控制之中,仿真计算结果表明这种算法加快了网络的收敛速度,收到了较好的解耦控制效果。
When using neural network to improve controlling parameter of PID controller,it has not a mature way to fix η and α.So this paper presents a fuzzy-number algorithm which can regulate η and α in real-time.After the integrated error quantity is processed by fuzzy quantification and normolized,η and α discover its corresponding subordination from the basis fuzzy subset membership function.Applying those to decoupling control of unlinear system,the simulation result indicated this algorithm speeds up convergence quickly and gets a better result of decoupling.
出处
《河南科技大学学报(自然科学版)》
CAS
2006年第6期50-53,共4页
Journal of Henan University of Science And Technology:Natural Science
关键词
神经网络
PID控制
参数优化
解耦控制
Neural network
PID control
Parameter improving
Decoupling control