LSE辅助变量选择算法及其在软测量中的应用

摘要

研究了基于人工神经网络的软测量技术在某炼油厂重整加氢装置预加氢汽油干点在线预测中的应用,提出一种基于最小均方预测误差(LSE)的辅助变量集选择算法.与PCA降维算法相比,LSE法能有效剔除冗余的辅助变量,降低软测量实施过程中的数据采集成本.%the anti-scaling rate were studied. Experimental results prove the microcrystal existence in solution because of different scale ion concentration between the filtered and unfiltered water; and the ultrasound can suppress conversion of scale ions to microcrystal, and the changing trend of microcrystal into deposit can be brought down and the fouling rate can be decreased. The double-ultrasound' s low anti-scaling rate than single-ultrasound's indicates that the ultrasound couldn't be superposed simply.