基于统计法的软测量建模研究

Research on Soft Sensor Modeling Based on Statistical Method

对比了机理法、统计法、机理统计混合法三种软测量建模方案,通过研究基于统计法的软测量建模方法及类型,提出了一种通用的软测量模型架构及其模型参数的整定方法。并以实际的苯乙烯精馏过程中的质量参数AMS浓度估计为实例,研究了预测偏差及过拟合问题,采用Kalman滤波对模型进行了更新,改善了模型预测效果,证明了算法的有效性,为后续的质量控制奠定了基础。

three soft sensor modeling schemes,including mechanism method,statistics method and mechanism statistical mixed method,are compared.by studying the modeling methods and types of soft sensor based on statistics,a general soft sensor model framework and its parameter setting method are proposed.Taking the AMS concentration estimation of the actual Styrene distillation process as an example,the problem of prediction bias and over fitting is studied.Kalman filter is used to update the model,the prediction effect of the model is improved,effectiveness of algorithm is proved,and a foundation for the subsequent quality control is laid.