基于神经网络过热汽温模型辨识

Model Identification of Superheater Temperature Based Neural Network

提出一种基于粗糙集理论的RBF网络构造算法。在RS-RBF算法中,通过粗糙集理论分析RBF网络隐层中的隐节点,将其分类,将不重要的隐节点删去。该方法能自动设计出规模合理的RBF网络结构,最终设计的网络不仅具有精简的网络结构,也具有较好的泛化能力。

In this article, a conformation method based rough sets of RBF （RS -RBF） neural networks is proposed. Using the principle of rough sets, the hidden units are class. The unimportant units will be pruned. So RS -RBF can produce a compact network structure automatically. Obtained RBF network can generalize well.