摘要
针对水泥熟料fCaO含量难以在线实时测量,提出了一种基于最小二乘支持向量机的软测量建模方法。针对最小二乘支持向量机模型的2个难点进行了改进:首先利用样本间的马氏距离来衡量样本的相似程度,删除样本中部分相似样本,提高最小二乘支持向量机模型的稀疏性,从而减小了模型的运算量。然后利用改进的粒子群优化算法对最小二乘支持向量机模型的2个重要参数进行迭代寻优,克服了常规交叉验证法或网格搜索法等参数选择方法的盲目性。最后将基于粒子群最小二乘支持向量机软测量模型用于熟料fCaO含量的实例仿真。结果表明,该方法具有收敛性好、预测精度高、泛化能力强等优点。
It is difficult to accurately measure the content of cement clinker fCaO in online and real-time way.In this study,a soft sensor modeling method based on least squares support vector machine （LSSVM） is provided.The following improvements are made to connect two difficulties of LSSVM model.Firstly,the Mahalanobis distance between samples is used to measure the similarity of samples,which is followed by deleting some similar samples and improving the sparsity of LSSVM to improve the computation model.Secondly,the improved particle swarm optimization （PSO）algorithm is used to iteratively optimize two important parameters of LSSVM model,which can overcome the blindness of parameter selection method of conventional cross validation method or the grid search method.Finally,the clinker fCaO content is simulated by means of the PSO LSSVM soft measurement model.It indicates that the method has good convergence,high precision and strong generalization ability.
出处
《现代化工》
CAS
CSCD
北大核心
2014年第6期152-155,共4页
Modern Chemical Industry
关键词
软测量
最小二乘支持向量机
稀疏性
粒子群优化
soft-sensing
least square support vector machine （LSSVM）
sparsity
particle swarm optimiziation （PSO）
