Studying Catalyst Activity in an Isomerization Plant to Upgrade the Octane Number of Gasoline by Using a Hybrid Artificial‐Neural‐Network Model

Abstract

AbstractIn this study, a hybrid model is presented for estimating the activity of a commercial Pt/zeolite catalyst in an industrial‐scale light naphtha isomerization unit. This model is also capable of predicting the research octane number (RON) and Reid vapor pressure (RVP) of gasoline, the flow rate of product, and the temperature profile of the reactor. In the proposed model, the decay function of heterogeneous catalysts is combined with a recurrent‐layer artificial neural network. To identify the activity of catalyst, a set of 52 data points during 272 days were gathered from the target isomerization plant. From these data, 31 points were selected for training (60 %), 11 data points for testing (20 %) and the remained ones for validating the developed hybrid network (20 %). Results show that the presented hybrid model can acceptably estimate the activity of the catalyst during its life in consideration of all process variables. Moreover, it is confirmed that for unseen (validating) data, the proposed model is capable of predicting RON, RVP, flow rate of gasoline, and the temperatures of the middle and end sections of the isomerization reactor with an average absolute deviations (AAD %) of 0.0769 %, 0.118 %, 0.6945 %, 0.042 %, and 0.1137 %, respectively.