摘要
时间序列预测技术可实现过程参数未来变化趋势的早期预报,从而为分析判断工况是否正常、确定转入下一工序的时机提供依据。针对间歇过程数据长度短、非线性、动态、不同批次数据不等长等特点,提出了一种基于相空间重构-最小二乘支持向量机的非线性时间序列预测方法。首先将多批次数据随机的拼接组成长数据向量,差分处理后采用相空间重构关联积分C-C方法计算该序列的延迟时间τ和嵌入维数m,从而构建训练集和检验集,然后采用最小二乘支持向量机算法建立预测模型。对某间歇蒸馏过程上升气温度建立的5步预测模型可用于生产现场的在线预报。
Time series prediction technology can be used to forecast process parameter change trends in the future, therefore this method can provide the basis for analyzing and deciding if process situation is normal, and when the process should switch to the next procedure. A nonlinear time series prediction method based on phase space reconstruction-least square support vector machine which is used to predict batch process variables that is nonlinear and dynamic, and whose length are short and unequal, is proposed. Firstly, batches process data is randomly linked to each other to build a long data series, and is differenced, and is used to compute delay time r and embedding variable number m using C-C method. Secondly, the data set, which is reconstructed according to delay coordinate method, is used to develop prediction model using least square support vector machine method. The developed five step prediction model to predict updraft temperature of one batch distillation process is proved to be precise and exact, and it can be used to predict online.
出处
《计算机与应用化学》
CAS
CSCD
北大核心
2011年第12期1577-1580,共4页
Computers and Applied Chemistry
关键词
间歇过程
相空间重构
最小二乘支持向量机
非线性时间序列预测
C-C方法
batch process, phase space reconstruction, least square support vector machine (LSSVM), nonlinear time series prediction, C-C method