密闭鼓风炉锌产量的支持向量机实时预报模型  

A Real-time Model for Forecasting Zinc Output by Support Vector Machining in Imperial Smelting Furnace

作　　者：胡志坤[1] 彭小奇[1] 桂卫华[2] 

机构地区：[1]中南大学物理科学与技术学院,长沙410083 [2]中南大学信息科学与工程学院,长沙410083

出　　处：《计算机工程》2004年第12期16-18,共3页Computer Engineering

基　　金：国家"973"计划基金资助项目(2002cb312200);国家自然科学基金资助项目(50374079);湖南省自然科学基金资助项目(01JJY2110)

摘　　要：为优化密闭鼓风炉的操作参数,建立了锌产量的实时预报模型。该模型采用分类SMO方法训练支持向量机回归模型,并根据若干步的误差来在线校正模型参数,对锌产量进行多步预报,以及时调整操作参数,并能在线学习预报模型。该预报模型的工业仿真表明在只有较少的样本数的情况下,在有效误差范围内能达到90%,且具有很好的实时性。该模型已应用于密闭鼓风炉操作优化与故障诊断系统,能较好地指导生产。A real-time model for forecasting zinc output by support vectors machine (SVM) is presented in order to optimize operational parameters of imperial smelting furnace (ISP). In this model, the learning method sequential minimal optimization (SMO) of the support vectors regression based on the support vectors for classification is adopted, the parameters of the model is adjusted by the errors of some steps, and the learning can be carried out on line. The industrial simulation results show that the practical forecast range of this model is up to 90%. This model has been applied to a system for optimization operation and fault diagnosis in ISP, and supervised production well.

关 键 词：支持向量机 回归 神经网络 密闭鼓风炉 锌产量 

分 类 号：TP182[自动化与计算机技术—控制理论与控制工程]