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.
出处
《计算机工程》
CAS
CSCD
北大核心
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)
关键词
支持向量机
回归
神经网络
密闭鼓风炉
锌产量
Support vectors machine
Regression
Neural network
Imperial smelting furnace
Zinc output
