Impact of Feed Composition and Operating Conditions on Product Yield and Quality, and Cycle Length of an Industrial Hybrid Naphtha Catalytic Reformer

Modern petroleum refining relies heavily on the performance of the catalytic reforming of naphtha, as its liquid product (reformate) provides about 30% of the refinery’s high-octane gasoline pool. The octane number of reformate, its yield, and the cycle length of the naphtha reformer significantly affect the overall petroleum refinery performance. This study examined an industrial hybrid reformer consisting of two interconnected units—a semi-regenerative reforming unit with three fixed-bed reactors and a continuous catalyst regeneration unit—capable of producing gasoline with a research octane number (RON) of 100 and motor octane number (MON) of 89. The objective was to quantitatively determine the influence of feedstock composition, catalyst regeneration, and operating conditions on product yield and quality, as well as on cycle length. Multiple linear regression (MLR) was used to calculate MON at time zero as a function of weight average initial temperature (WAIT), feed rate, and feed composition. Catalyst deactivation based on reformate MON, calculated by the MLR, was modelled using nonlinear least-squares minimization and bi-population covariance matrix adaptation evolution strategy (BIPOP-CMA-ES). Feedforward deep neural network (FFDNN) models with three hidden layers, containing 128 neurons, were developed for the reformate yields and octane number. The mean squared error of the FFDNN model of MON for the unseen data set was 0.07, compared with 0.45 for the MLR. The regeneration of the catalyst showed that it deteriorated in activity and stability, most probably due to the loss of rhenium content observed during the regeneration process.