基于FCM的分布式RBF网络模型对高炉炉温的预测研究

Study on Temperature Prediction in Blast Furnace by Using Distributed RBF Neural Network Model Based on FCM

针对高炉炉温单模型难以预测的问题。根据分布式建模的思想,提出一种基于FCM的分布式神经网络模型,先用模糊C均值聚类对输入输出样本空间进行模糊划分,然后对每个划分建立RBF神经网络子模型,使用子模型对测试样本集的样本点进行预测,以测试样本点对每一划分的隶属度为权值,进行加权求和,得到最终预测值。研究表明,对比单一神经网络模型,基于FCM的分布式RBF网络模型有更好的预测效果。

According to the ideas of distributed modeling,this paper proposes a kind of distributed neural network model based on FCM in order to solve the problem that a single model is difficult to predict the temperature of blast furnace.Firstly,divide the sample space of the input and output into multiple sub-space by using fuzzy C-means clustering.Then establish RBF neural network sub-net model for each subspace,then use the sub-net model to predict each sample points of the testing set.The membership of each point and each sub-space is thought of as the weight value.Use the average weighted sum to obtain the final predicted value.