# Catalysis 2022 issue

cm 2 , during the period of August to November 2020 when there was change in operating conditions, especially in terms of RON. A sum- mary of average RON during the last 10 years is illustrated in Table 1 . Objectives The study’s objective is to identify the determinants and explanatory variables affecting the behaviour of the first reformer reactor ∆P to control it and avoid any emergency shutdown and its reoccurrence. Determining factors affecting ∆P in first reactor There were several variables affect- ing the behaviour of the ∆P in the first reactor. A total of 10 variables were identified then each variable was tested for its significance The basic method is to establish the linear relationship between the output (first reformer reactor ∆P) and the influencing or explanatory variable. The identified variables for the regression analysis are: Heater inlet temperature Reactor inlet temperature RON control temperature Unit feed Recycle gas flow Hydrogen to hydrocarbon ratio Hydrogen flow to first reactor for elutriation From the basic multiple lin- ear regression (MLR) equation, y = xβ, the basic form MLR can be expressed as: y=β 0 + β 1 x 1 + β 2 x 2 + β 3 x 3 + β 4 x 4 + β 5 x 5 + β 6 x 6 + β 7 x 7 + β 8 x 8 + β 9 x 9 + β 10 x 10 Total lift gas flow Secondary lift gas Lift “A” partial ∆P Initially, 10 variables were iden- tified for the MLR. A list of data for these variables from 01/01/2020 to 22/10/2020 was used to conduct the analysis and calculations utilis- ing Microsoft Excel and statistical software based on MLR modelling where: y = First reformer reactor ∆P (response variable) x 1 = Total lift gas flow

x 2 = Secondary lift gas x 3 = Lift “A” partial ∆P x 4 = Heater inlet temperature x 5 = Reactor inlet temperature x 6 = RON control temperature x 7 = Unit feed x 8 = Recycle gas flow x 9 = Hydrogen to hydrocarbon ratio x 10 = Hydrogen flow to first upper hopper for elutriation The linear regression method was used to calculate the regression coef- ficients with 10 independent var- iables. The regression coefficients from β 0 to β1 0 are respectively given as: β 0 = -0.933 β 1 = -0.003578

the significance requirements of α = 0.05. After screening the P-value and contrast behaviour variables, the below variables have been consid- ered for final MLR with the response variables of the first reformer reactor ∆P: x 4 = Heater inlet temperature x 6 = RON control temperature x 7 = Unit feed x 8 = Recycle gas flow x 9 = Hydrogen to hydrocarbon ratio x 10 = Hydrogen flow to first upper hopper for elutriation Therefore, a new mathematical model is obtained and the regression equation can be rewritten as:

β 2 = -30.3 β 3 = 29.5

y = -1.413 + 0.00670x 4 - 0.00393x 6 - 0.004184x 7 + 0.01202x 8 - 0.1775x 9 + 0.001445x 10

β 4 = 0.00343 β 5 = 0.02986 β 6 = -0.03099 β 7 = -0.004933 β 8 = 0.01719 β 9 = -0.1847 β 10 = 0.001375 The

Discussion As it can be seen, the RON control temperature is the most significant explanatory variable in relation to the increase in ∆P. The increase in RON control temperature results in an increase in the first reformer reactor ∆P; it can be observed that there is a linear relationship between the increase in RON control tem- perature and the first reformer reactor ∆P. The behaviour of explan- atory variable x 6 = RON Control and response variable y = first reformer reactor ∆P is illustrated in Figure 2 . The refiner’s attention was drawn to the fact that the increase in first reformer reactor ∆P will show days or weeks after operating with higher RON – from RON 95 to RON 98 – after which the situation cannot be improved, which was the case in the refinery. As the severity increased, either as the RON increased from 95 to 98, or as the RON control temperature increased (as shown by an increase in weighted average inlet tempera- ture (WAIT)), there was an increase in dehydrocyclisation of paraffins to aromatics as well as increased crack- ing and coking, decreased reformate yield, and decreased H 2 purity. This led to higher catalyst attrition (i.e. broken particles, catalyst fines and Heater inlet temperature/RON control temperature

mathematical

regression

model is obtained as:

y = -0.933 - 0.003578x 1 - 30.3x 2 + 29.5x 3 + 0.00343x 4 + 0.02986x 5 - 0.03099x 6 - 0.004933x 7 + 0.01719x 8 - 0.1847x 9 + 0.001375x 10 The less significant variables were rejected one by one based on the P-value. The significant indicator α = 0.05 is considered as the screening index. When the P-value > 0.05, the least significant variable should be removed. Otherwise, the result is the opposite. For instance, the P-value for Lift “A” partial ∆P is 0.223 in the first-round screening and P-value > 0.05; therefore, this item should be rejected. In addition, the variables that show a constant behaviour, or those which another parameter can explain, were eliminated. Finally, the rejected variables are: x 1 = Total lift gas flow x 2 = Secondary lift gas x 3 = Lift “A” partial ∆P x 5 = Reactor inlet temperature (as it can be explained by x 6 ) From the calculation, the P-value of all remaining variables satisfies

30 Catalysis 2022

www.digitalrefining.com