一种动态校正的AGMM-GPR多模型软测量建模方法

发布时间：2018-11-29 08:40

【摘要】：工业过程常常是强非线性的,并有多个工况,传统的软测量方法存在预测能力差,不能有效利用误差信息等缺点.为了有效解决这些问题,提出一种基于自适应高斯混合模型-高斯过程回归(AGMM-GPR)的多模型动态校正软测量建模方法.首先,通过贝叶斯信息准则构建自适应高斯混合模型(AGMM),得到优化的子模型个数;然后,利用GPR方法建立各局部模型,当新的数据到来时,将其隶属于各局部模型的后验概率和预测值融合得到多模型输出;最后,为了进一步提高模型的精度,构建自回归积分滑动平均(ARIMA)模型对多模型输出进行动态反馈校正.通过数值仿真和硫回收装置(SRU)中H2S浓度的估计,验证了所提方法具有良好的预测精度和泛化性能.
[Abstract]:Industrial processes are often strongly nonlinear and have multiple operating conditions. The traditional soft sensing methods have some disadvantages such as poor prediction ability and poor utilization of error information. In order to solve these problems effectively, a multi-model dynamic correction soft sensor modeling method based on adaptive Gao Si hybrid model-Gao Si process regression (AGMM-GPR) is proposed. Firstly, the optimal number of sub-models is obtained by constructing adaptive Gao Si hybrid model (AGMM), by Bayesian information criterion. Then, each local model is established by using GPR method. When the new data comes, the local model output is obtained by merging the posterior probability and prediction value of each local model. Finally, in order to further improve the accuracy of the model, an autoregressive integral moving average (ARIMA) model is constructed for dynamic feedback correction of multiple model outputs. By numerical simulation and estimation of H _ 2S concentration in (SRU) of sulfur recovery unit, it is proved that the proposed method has good prediction accuracy and generalization performance.
【作者单位】：江南大学轻工过程先进控制教育部重点实验室;江南大学物联网工程学院自动化研究所;
【基金】：国家自然科学基金资助项目(21206053;21276111) 江苏省“六大人才高峰”计划资助项目(2013-DZXX-043) 江苏省产学研资助项目(BY2014023-27) 江苏高校优势学科建设工程资助项目(PAPD)
【分类号】：TP274

【参考文献】