基于混合学习算法的RBF神经网络主蒸汽温度控制  

MAIN STEAM TEMPERATURE CONTROL USING RBF NEURAL NETWORK BASED ON HYBRID LEARNING ALGORITHM

作　　者：王杰[1] 姜国强[1] 王栓[1] 

机构地区：[1]郑州大学电气工程学院,河南郑州450001

出　　处：《热力发电》2009年第2期28-31,36,共5页Thermal Power Generation

基　　金：河南省杰出人才创新基金项目(074200510013);河南省教育厅自然科学基金项目(2007520048)。

摘　　要：针对火电厂主蒸汽温度的大迟延、模型不确定性特点,提出一种使用径向基(RBF)神经网络整定PID串级主蒸汽温度控制策略。采用一种最近邻聚类法和梯度下降法相结合的混合学习算法构造RBF神经网络,在线辨识被控对象并对PID主控制器参数进行在线调整。仿真结果表明,基于混合学习算法的RBF神经网络PID控制器具有控制精度高、响应速度快的优点,系统动态品质优于常规算法的RBF神经网络PID控制。Directing against the features of long time delay and uncertainty in the main steam temperature control of thermal power plants, a RBF -PID cascade control strategy has been put forward. Adopting a hybrid learnging algorithm, which combines the nearest neighbor- clustering method with the gradient descent method, to construct a RBF neural network, the controlled objects can be iden- tified on - line and the parameters of PID master controller can be adjusted on - line. The simulation results show that the PID control- ler using RBF neural network based on hybrid learning algorithm has the advantages of high precision control and rapid response speed,the dynamic quality of system being more better than that of PID controller using RBF neural network of conventional learning algorithm.

关 键 词：火电厂 主蒸汽温度 控制 最近邻聚类法 梯度下降法 混合学习算法 RBF神经网络 PID 

分 类 号：TP273.2[自动化与计算机技术—检测技术与自动化装置]