摘要
针对一般的神经网络PID控制器难于得到系统预测输出值的缺陷,提出一种改进型的粗糙集神经网络PID控制器,阐述了其设计原理;对串级过热汽温控制系统的仿真结果表明,与普通神经网络PID建模方法相比,该控制器抗干扰和鲁棒性强。
A new improved RS-based neural PID controller is introduced in this paper.As It is difficult for RS-based neural PID controller to obtain the predicted output of the system, aiming for this defect, the new controller uses predicted mathematical model instead of actual output,and reduces redundant attributes and the values of a stylebook in BP network to keep important information by rough set;and the least stylebook is acquired which trains the neural network. Simulation results of the super heated steam temperature cascade control system show that the resistance and robustness of the system are good.
出处
《太原理工大学学报》
CAS
2004年第3期251-253,共3页
Journal of Taiyuan University of Technology
基金
国家自然科学基金资助项目(60374029)
山西省留学回国人员基金资助项目(200027)
关键词
粗糙集
数据约简
BP网络
建模
rough set
data reduction
BP network
modeling
