Simulation of the physical temperature probes for soft sensor design under feed composition changes for naphtha plant


, Q2


This paper presents a versatile strategy for using simulations to identify control stages of naphtha distillation plant where additional temperature probes can be installed. This search is based on sensitivity analysis applied to the real distillation columns under different operating modes, while simulating the data probed with added thermocouples. It was shown that feed composition has a crucial influence on the resulting position of the control stage(s) and efficiency of the soft sensor (SS) in predicting the quality product. In most practical cases, when the oil field changes, the accuracy of the SS decreases significantly and requires rebuilding, which poses a problem in the control of the technological process. As shown in the paper, SS built with additional thermocouples used as inputs has demonstrated increase in the prediction accuracy under different feed compositions. The effectiveness of the suggested strategy is confirmed by comparing the results with those assuming the use of only physically available probes as inputs for SS.