基于LS-SVM的煤气发生炉关键参数预测  

Parameters predicting model of gas furnace based on least squares support vector machine

作　　者：蒋少华[1] 

机构地区：[1]韶关学院计算机科学学院,广东韶关512024

出　　处：《中南大学学报（自然科学版）》2014年第6期1873-1878,共6页Journal of Central South University:Science and Technology

基　　金：国家自然科学基金资助项目(61101134);韶关市科技计划项目(韶科(管)[2010]-11)

摘　　要：根据煤气炉现场采集的数据,建立一种基于最小二乘向量机(LS-SVM)的预测模型,预测煤气炉关键参数炉出温度、CO2含量。模型以主要工艺参数作为影响因素,以炉出温度、CO2含量为影响对象,建立影响因素和影响对象之间的复杂非线性关系,构造煤气炉参数LS-SVM预测模型,再运用奇异值分解的方法辨识模型参数,最后将模型用于煤气炉参数预测。研究结果表明:该模型能及时跟踪炉况参数的变化,预测结果与实测值较吻合,准确度与处理速度都优于神经网络预测模型,实际预测误差小于2%,可用于煤气炉生产过程的现场操作指导。The gas furnace temperature and CO2 content in gas are the key parameters which reflect whether the furnace condition is normal. In order to keep the gas furnace （GF） working smoothly, a model based on least squares support vector machine （LS-SVM） was presented. With the main data samples as influence factors, and with the furnace temperature and CO2 content in gas as influence object, the complex nonlinear relations among the influence factors and influence objects were fitted by LS-SVM model. Firstly, the predicting model was constructed, and then a numerical algorithm for subspace system （singular value decomposition, SVD） was utilized to identify the model. Finally, the model was used to predict the furnace parameters. The results show that the prediction accuracy and treatment speed by this model are much higher than those of back-propagation neural networks（BPNN）, and the practical prediction errors are less than 2.0%.The monitoring model is applied in the assistant decision-making system of a gas furnace.

关 键 词：最小二乘向量机(LS-SVM) 参数预测 煤气发生炉 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]