摘要
根据某企业德士古气化炉装置在线估计炉温的需要,将现场数据采样样本中的离群点分为高杠杆点和高残差点两类,将一种新的加权方法应用到最小二乘支持向量机(LS-SVM),使其对两种离群点都具有抑制作用,提高模型鲁棒性。加权最小二乘支持向量机(Weighted LeastSquare Support Vector Machine,WLS-SVM))参数的选择基于LS-SVM的最优参数,根据模型训练误差对参数进行二次寻优,进一步提高模型精度。利用测试函数验证了改进方法,对提高模型精度有明显效果;并将改进方法应用到实际生产装置的炉温软测量系统中,也取得了满意的应用效果。
To satisfy the requirement of the online estimation of temperature for Texaco slurry gasifier,the outliers of the sampled data from Texaco slurry gasifier are divided into two categories,i.e.,vertical outlier and leverage point.A new weighted least square support vector machine（WLS-SVM） is proposed to attenuate the two kinds of outliers.The parameters in WLS-SVM are decided by the optimal parameters of LS-SVM,which are further optimized by means of the generalization error of LS-SVM model.Simulation experiments based on test functions illustrates the effectiveness of the proposed method.Finally,the present method is applied to the soft sensor model for the temperature of Texaco slurry gasifier,and some satisfying results are obtained.
出处
《华东理工大学学报（自然科学版）》
CAS
CSCD
北大核心
2010年第5期717-722,共6页
Journal of East China University of Science and Technology
关键词
软测量
德士古炉温
加权最小二乘支持向量机
参数二次优化
soft sensor
temperature of Texaco slurry gasifier
weighted LS-SVM
re-optimize super-parameters
