关键词: 径向基函数, 神经网络预测, 全局最优, 在线学习, 动态优化

Abstract: This paper presents a RBF（Redial Basis Function） neural network controller on superheat temperature system in supercritical units, a sequential algorithm for RBF networks referred to as the generalized growing and pruning algorithm for RBF (GGAPRBF) is introduced and then uses it in the learning algorithm to realize parsimonious networks. The structure of this controller makes no need to use another neural network for online system identification and determining the structure of neural network controller in advance．The simulation for Super Heat Temperature control system using presented method is take out．The results show that the control system performance is better than the conventional PID control system．

Key words: radial basis function, neural network prediction, global approximation, online learning, dynamic optimize