最小RBF网设计的进化优选算法及其在动力配煤过程状态预测建模中的应用

文中提出一种RBF网设计的进化优选算法，该算法能设计出满足给定学习精度的最小RBF网络，从而使所得到的RBF网具有较好的泛化能力。我们以煤灰软化温度预测为例，介绍了如何将该算法应用于动力配煤过程的状态预测建模。我们还把ESA算法与另外2种最常见的RBF网设计方法，即聚类方法和OLS算法作了比较，结果表明，用进化优选算法设计的RBF网模型具有最好的泛化能力。

EVOLUTIONARY SELECTING ALGORITHM FOR DESIGNING MINIMAL RBF NETS AND ITS APPLICATION IN COAL BLENDING

This paper proposes an evolutionary selecting algorithm (ESA) to design RBF nets. Given an accuracy limit, ESA can design a RBF net with east hidden units, so the network will generalize well. Proposed algorithm can be used to establish models of state estimation process for steam coal blending. One of such modeling problems, softening temperature prediction of coal ash, is illustrated using ESA method. Minimal RBF net designed is compared with RBF nets obtained with OLS and clustering method, result shows that RBF net designed with ESA method has better generalization ability.