基于神经网络的自整定PID控制器设计  

Design of Self-tuning PID Controller Based on Neural Network

作　　者：张世峰[1] 李鹏[1,2] 

机构地区：[1]安徽工业大学电气信息学院,安徽马鞍山243002 [2]日照钢铁有限公司第二炼钢厂,山东日照276806

出　　处：《自动化仪表》2009年第7期64-66,共3页Process Automation Instrumentation

基　　金：安徽省教育厅自然科学研究基金资助项目(编号:KJ2008B104)

摘　　要：针对非线性时变系统,设计了一种基于神经网络的参数在线自整定PID控制器。该控制器采用基于最近邻聚类方法的RBF神经网络快速学习算法,通过实时在线辨识,建立被控系统的精确模型并得到准确的Jacobian信息;同时将此信息提供给BP神经网络,从而实现PID控制器参数的自动在线整定。仿真结果表明,该方法提高了算法的精度和速度并具有较快的系统响应和良好的跟踪特性。For nonlinear time varying system, a parameters self-tuning PID controller based on neural network is designed. In the controller, the nearest neighbor clustering based RBF neural network fast speed learning algorithm is adopted, through real-time online identification, proper model of the controlled system is established, and precise Jacobian information is obtained. In addition, the information is provided to BP NN, thus automatic online tuning for parameters of the PID controller is implemented. The result of simulation proves that the method enhances both accuracy and speed of the algorithm; the controller offers fast response and good tracking characteristics.

关 键 词：BP神经网络 RBF神经网络 最近邻聚类 PID控制 参数自整定 

分 类 号：TP273.2[自动化与计算机技术—检测技术与自动化装置]