摘要
提出了一种具有强非线性表达能力的自适应偏最小二乘回归(APLSR)方法,并应用于初顶石脑油干点软测量模型建立.APLSR对于指定的预测对象,将根据样本在自变量空间中的分布,分析它们对预测对象的预报能力,自适应地为各个样本分配权值,然后从加权样本数据中提取和选定PLS成分,实施自适应加权PLSR,从而获得预报性能良好的模型.同时提出将前一时刻初顶石脑油干点人工分析值引入作为模型的自变量,从而进一步提高了软测量模型的预测精度.
A novel adapting partial least square regression （APLSR） approach was proposed to develop the naphtha dry point soft sensor of the primary distillation tower. Many operation conditions were related to naphtha dry point and there existed a significant correlation among them. In order to obtain a naphtha dry point soft sensor with high predicting correctness, the different predicting contribution ratios of modeling samples were taken into account by APLSR and the optimal number of the latent variables was obtained according to the predicting ability of the soft sensor. When APLSR was used for the predicting sample, each modeling sample was weighted according to its ratio of predicting contribution for the predicting sample and satisfactory results were obtained. Further, the previous analysis value of the naphtha dry point was regarded as a new independent variable for the soft sensor and the predicting correctness of the soft sensor was enhanced remarkably.
出处
《化工学报》
EI
CAS
CSCD
北大核心
2005年第8期1511-1515,共5页
CIESC Journal
基金
上海启明星项目(04QMX1433)
国家重点基础研究发展规划项目(2002CB312200)
国家高技术研究发展计划项目(2002AA412110)~~
关键词
自适应
加权回归
偏最小二乘回归
石脑油
干点
软测量
adapting
weighted regression
partial least square regression
naphtha
dry point
soft sensor
