基于FCM的多支持向量机模型在高炉炉温预测中的应用  

Application of multiple support vector machines model based on FCM for temperature prediction in blast furnace

作　　者：王义康[1,2] 刘祥官[2] 

机构地区：[1]中国计量学院理学院,浙江杭州310018 [2]浙江大学

出　　处：《冶金自动化》2012年第3期18-23,共6页Metallurgical Industry Automation

基　　金：国家自然科学基金资助项目(60911130510;61101239);浙江省科技厅省级公益性技术应用研究计划资助项目(2011C21020)

摘　　要：针对高炉炼铁智能控制专家系统中单一支持向量机(SVM)炉温预测模型的改进研究,提出一种基于模糊C均值聚类(FCM)的多支持向量机模型。首先运用模糊C均值聚类对模型训练集进行聚类划分,然后对每一类进行支持向量机的训练,建立相应的子模型,并对测试集中的同一样本点分别进行预测,以测试样本点的输入对应于每一类的隶属度为权值,进行加权求和,最终得到预测值。通过对在线采集的数据分析表明,基于FCM的多支持向量机模型比单一的支持向量机模型在多方面预测性能得到改善,连续预测100炉命中率达86%。A kind of multiple support vector machines model based on fuzzy C-mean clustering （FCM） is proposed in order to improve the temperature prediction model with single support vector machine （SVM） in intelligent control expert system for blast furnace. At first, the training set is clas- sified into several groups by using of fuzzy C-mean clustering method. Then each group is trained through support vector machine and the corresponding sub-models are established. After that, the same sample in testing set is predicted separately and the memberships between each sample and each group are set as the weights. Finally, the predicted result is obtained by weighted sum. The analysis of on-line data shows that the multiple support vector machines model based on fuzzy C-mean clustering has. better prediction performance in many aspects compared with the single SVM model. The hit rate reaches 86% in consecutive 100 furnaces based on the proposed model.

关 键 词：高炉 炉温预测 模糊C均值聚类 支持向量机 

分 类 号：TF543[冶金工程—钢铁冶金]