摘要
提出了一种基于滚动时间窗的最小二乘支持向量机(LSSVM)回归估计方法.该方法构造了滚动时间窗,利用滚动时间窗内的数据优化建模.模型随着时间窗的滚动进行在线更新,并对滚动时间窗内的数据分配不同的权值以充分利用数据的信息.将基于滚动时间窗的LSSVM回归估计方法应用于软测量建模,进行轻柴油凝固点的预估.结果表明,该建模方法十分有效.
The least square SVM regression method based on sliding time window. A sliding time window is built and its data are employed to construct the dynamic model of a system. This model is refreshed on-line with the rolling of the time window. Different weights are assigned to the data in sliding time window in order to richly exploit the data information. The method was employed to construct soft sensor model, which is applied to estimate the frozen point of light diesel oil.
出处
《上海交通大学学报》
EI
CAS
CSCD
北大核心
2004年第4期524-526,532,共4页
Journal of Shanghai Jiaotong University
基金
国家高技术研究发展计划(863)资助项目(2001AA413130)
关键词
最小二乘支持向量机
软测量
滚动时间窗
建模
Computer simulation
Estimation
Least squares approximations
Regression analysis
Supports
Vectors
