Soft sensors based on dynamic PLS (DPLS) have been widely used in industrial applications for predicting hard-to-measure quality variables. However, DPLS is prone to overfitting due to an increasing number of model inputs. A plethora of approaches have been proposed to improve DPLS-based soft sensors, among which variable selection has been a prevailing one. Recently, a new method termed as DPLS-TS has been proposed to penalize dynamic parameters in DPLS using a temporal smoothness regularization, which helps reduce model complexity and deliver smooth predictions for quality variables. In this work we present a comparative study of temporal smoothness regularization and variable selection in terms of their improvements in prediction performance when a large number of lagged time series data are involved. Comparisons are performed through a simulated case of crude distillation unit.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.