基于M-SVR与RVFLNs的高炉十字测温中心温度估计  

Centre Temperature Estimation of Blast Furnace Cross Temperature Measuring Based on M-SVR and RVFLNs

作　　者：周平[1] 尤磊[1] 刘记平 张兴[1] 

机构地区：[1]东北大学流程工业综合自动化国家重点实验室,辽宁沈阳110819

出　　处：《东北大学学报（自然科学版）》2017年第5期614-619,共6页Journal of Northeastern University(Natural Science)

基　　金：国家自然科学基金资助项目(61473064;61290323;61333007;61290321;61621004);中央高校基本科研业务费专项资金资助项目(N160805001;N160801001);辽宁省教育厅科技项目(L20150186)

摘　　要：由于高炉中心温度较高,十字测温中心位置传感器极易损坏,并且更换周期长,因而导致无法及时判断炉顶煤气流分布.采用多输出支持向量回归(M-SVR)和随机权神经网络(RVFLNs)两种数据驱动智能建模方法建立高炉十字测温中心带温度估计模型,并基于实际工业数据对建立的模型进行验证和比较分析.结果表明,在样本数量较小时,M-SVR模型和RVFLNs模型都具有较好的温度估计效果,但当样本数量充足时,M-SVR模型的泛化性能和估计精度更优于RVFLNs模型.Due to the high temperature in the middle of blast furnace, the central position sensor of the cross temperature measuring is very easy to be damaged, and the replacement period is always long, resulting in the gas flow distribution not being observed in time. To this end, two kinds of data： based intelligent modeling methods of multi-output support vector regression machine （ M - SVR） and random vector functional-link networks （ RVFLNs） were used to establish the temperature estimation model of cross temperature measuring center of blast furnace. Finally, the temperature estimation model based on industrial data was verified and compared. The results show that both M - SVR model and RVFLNs model have good temperature estimation effect when the sample size is small. However, when the sample size is large enough, the generalization performance and estimation accuracy of M - SVR model is better than those of the RVFLNs model.

关 键 词：高炉炼铁 十字测温 温度估计 多输出支持向量回归机 随机权神经网络 

分 类 号：TP182[自动化与计算机技术—控制理论与控制工程]