RBF网络在线辨识算法与应用研究

A study of on-line identification algorithm by using RBF network and its application

提出了一种确定RBF网络中心向量的动态递推算法 ,介绍了RBF网络线性层连接权值的递推最小二乘算法 ,利用这些算法对蒸发器过热度进行了在线辨识 ,同时对网络结构和算法的参数变化对辨识精度的影响进行了研究 .仿真结果表明该算法辨识速度快、精度高 ,可用于在线控制 .

A dynamic recursive algorithm of RBF network center vector is proposed, and a recursive least squares algorithm for linking weight between linear layers of RBF network is introduced. The superheat degree in an evaporator is identified on line by using this algorithm. Meanwhile, the influence of the network structure and the change of the parameters in the algorithm on the precision is studied, also. The simulation results show that the algorithm can be used in on line control with fast speed and high accuracy of identification.