S HORTCUT A PPROACH T O M ODELING AND S IMULATION O F A C ATALYTIC R EFORMING P ROCESS

by R. M. Ansari and M. O. Tad Ansari holds masters degrees in chemical engineering, petroleum refining engineering and applied mathematics. He is senior process control engineer, Riyadh Refinery, Engineering Division. He has 15 years professional engineering experience and has published 25 refereed articles in professional journals. Tad holds a doctorate in chemical engineering from Queens University Canada and is an associate professor.

ABSTRACT
There are a number of different methods available to model the catalytic reforming process, and the main idea behind all these models is to determine the operating conditions of the reforming unit and to predict yield of reformate and reactor temperature profile accurately. Simulation of the catalytic reforming process is carried out using the Smiths kinetic scheme. The model results were tested in real-time application on an existing plant, and the results were reviewed to improve the processing conditions. The process model of the catalytic reforming process was also used in multivariable control application to provide target values for the reactor inlet temperatures. The need for looking into the simple model was due to the complexity of other models that accompany multicomponent reactions. Also the component data for specific rate constant, pressure exponent and equilibrium constant for several of these reforming reactions are not easily available in literature. The simple kinetic scheme as compared to the other complex models described in literature provides a shortcut approach for estimating the product yield of the catalytic reforming process.

Reactors / Heaters

Naphtha Feed

Feed Exchangers

Recycle Gas Excess Gas Gas & LPG

Separator Debutanizer

Platformate

Fig. 1. Schematic of catalytic reforming process

INTRODUCTION
The catalytic reforming process is used to increase low-octane naphtha fractions to high-octane reformate for use as a premium motor fuel blending component or as a source of aromatic hydrocarbons. Although a number of reactions take place during reforming, the predominant is the dehydrogenation of naphthenes to form aromatics. Some of these aromatics are isolated to become petrochemical feedstocks, but most become motor fuel blending stock of high antiknock quality. Most feedstocks for reforming are hydrotreated first to remove arsenic, sulfur and nitrogen compounds. Otherwise these compounds would poison the reforming catalyst. The cost of the hydrotreating step is easily justified by extending the life of the reforming catalyst. Hydrogen is a by-product of catalytic reforming. Some of this hydrogen is recycled to sustain reformer reactor pressure and to suppress coke formation. A schematic of the catalytic reforming process is shown in fig. 1. Feed to the reactors is combined with recycle hydrogen and preheated with reactor effluent. The feed is then brought up to reaction temperature in a direct-fired heater before entering the first of four reactors arranged in series. Since the net reforming reactions are endothermic, additional heat must be provided to the effluent from one reactor before entering the next. This is accomplished by direct-fired heaters on the inlet to the subsequent

three reactors. Temperature controls on the outlet of each of the heaters provides the basis for controlling heat input to each of the reactors. Modeling and simulation work on the catalytic reforming process is carried out using Smiths kinetic scheme (Smith 1970), and the model results were tested on an existing plant. The results from the dynamic model were reviewed to improve the processing conditions, and the same model was used within the catalytic reforming multivariable control applications. Smiths kinetic scheme was used to develop a dynamic model for multivariable control applications because it is simple to develop a dynamic model for control purposes from that scheme as compared to other complex kinetic schemes available in literature. The simple kinetic scheme provides a shortcut approach for estimating the product yield of the catalytic reforming process.

MODEL DEVELOPMENT
There are a number of different methods for modeling the catalytic reforming process, and the main idea behind all these models is to determine the operating conditions of the reforming unit and to predict the yield of reformate and the temperature profile accurately. A model must predict behavior not only within the reactor, but in the auxiliary areas of the unit as well. It should also consider the complex nature of the process and the reactions which take place during the process of reforming.

14

SAUDI ARAMCO JOURNAL OF TECHNOLOGY SPRING 1998

Other kinetic schemes available in literature are very complex, as these schemes consider secondary reactions taking place in the catalytic reforming. From a fundamental point of view, kinetics contribute to a better understanding of the reaction mechanisms and of the effect of catalyst upon these. From a practical point of view, accurate kinetic equations are of crucial importance for the reliable design and simulation of reactors. Kinetic analysis of reforming process is still following very complex lines. Multicomponent reactions accompanying catalytic reforming provide an insight into the more complex relations of a commercial operation, but the component data for specific rate constant, pressure exponent and equilibrium constant for several of reforming reactions is not available in literature. A mathematical model was developed by assembling the mass and energy balances on the system of reactions. The mass balance provides the variation of concentration of the components selected along the reactors, and the energy balance gives the variation of temperature. Mass and energy balances were carried out on one element of radial section of the reactor and then integrated over the whole reactor. Calculation of a reaction rate for all of the feed components over several reactors with declining temperature is the heart of establishing a model. The following is the kinetic scheme proposed by Smith (1970) for the process of catalytic reforming. n k5 k1/k2 Gas Naphthene + H2 Aromatics + 3H2 3 k3k4 n k6 Gas Paraffin 3

dT 1 = [(H1)(r1 r2) + (H2)(r4 r3) dw GCP + (H3)r5 + (H4)r6] (6)

From equations 2 to 6, r1...r6 are defined as the rate constants; GB and MB are the mass flow rates of naphtha and recycle gas; YN, YP, YA and YG are the mole fractions of naphthenes, paraffins, aromatics and gases, and n denotes the average number of carbon atom in molecules. The rate constants and their dependence on temperature were obtained from Smith (1970). All the main reactions are reversible, and their corresponding equilibrium constants were obtained from basic thermodynamic properties using the following: F = RT In K (7)

For other reactions, however, it was necessary to calculate F as follows:

T H dT F = T T2 T0

(8)

where H is the standard enthalpy change and F is the change in free energy associated with a particular reaction at a set temperature.

NUMERICAL INTEGRATION
The system of first-order differential equations was solved by the method of Runge-Kutta. The Runge-Kutta numerical integration technique improves upon the accuracy of a simple integration by means of interpolation in which the nonlinearity of the equation is compensated. PONA (paraffin, olefin, naphthene and aromatics) analysis gives the concentration of components on percent volume; these were converted to mole percentage.

(1)

Taking material balances on all the four components and energy balances on a differential element, dw, of catalyst, the following system of differential equations was obtained: MB dyN = ( r1 + r2 r4 + r3 r5) dw GB

RESULTS AND DISCUSSION

(2) The test runs were carried out on a catalytic reforming unit, and the results predicted by the model were compared with the plant data. Table 1 compares the overall results of reforming obtained by the model and the plant test runs. A number of model validations were needed through many sets of real process data to get these results. It may be observed from the table 1 that the model results differ with the plant test data for reformate and light gases. In this process model, this was not considered significant, as the main important variables were reactor temperatures and octane number. It may also be observed from the model results that low-pressure reforming has an increasing effect on the production of aromatics. The operation at low pressure increases the fraction of feed energy, in terms of gross heat of combustion, contained in the desired product, reformate and hydrogen.

MB dyP = ( r3 + r4 r6) dw GB

(3)

MB dyA = (r r ) dw GB 1 2

(4)

MB n n dyG = ( r5 + r6) dw GB 3 3

(5)

SAUDI ARAMCO JOURNAL OF TECHNOLOGY SPRING 1998

15

TABLE 1. PLANT TEST RESULTS COMPARED TO MODEL YIELDS Variables Outlet Temperature, Reactor (R1) Outlet Temperature, Reactor (R2) Outlet Temperature, Reactor (R3) Outlet Temperature, Reactor (R4) Yield of Reformate C5+, Vol % Production of Gases, Vol % Hydrogen C1 C2 C3 C4 TOTAL Composition of C5+, Vol % Paraffins Naphthenes Aromatics RON, Clear Model Results 450C 464C 485C 492C 89.60 Plant Test Results 446C 461C 486C 493C 85.34

1.90 1.10 1.89 2.10 2.45 99.04

2.75 1.25 2.23 4.36 4.02 99.95

45.30 1.30 53.30 97.80

47.52 1.74 50.00 97.50

TABLE 2. THE PREDICTION OF PRODUCT YIELDS AND REACTOR OUTLET TEMPERATURES FOR CASE 1: HIGHER NAPHTHA FEED RATE Temperatures And Yields Outlet Temperature, Reactor (R1) Outlet Temperature, Reactor (R2) Outlet Temperature, Reactor (R3) Outlet Temperature, Reactor (R4) Yield of Reformate C5+, Vol % Production of Gases, Vol % Hydrogen LPG Model Results 452C 470C 498C 517C 82.60 Plant Test 450C 468C 495C 510C 86.00 Relative Error 0.4% 0.4% 0.6% 1.4% 4.1%

The process model was used to predict the product yields and operating conditions under different reactor inlet temperatures, feed changes and naphtha feed compositions. Two cases were studied at different operating conditions, and the results are given in tables 2 and 3. These tables show that the maximum relative errors in the reactor outlet temperatures and product yields between the model prediction and plant were 1.4% and 5.5%, respectively. At higher feed rates (higher liquid hourly space velocity), fewer catalyst sites will be available per unit of feed, and so higher reactor temperatures will be required for the same conversion (see table 2). The net effect of an increase in feedrate at constant octane will be an improved platformate and hydrogen yield. Conversely, low throughputs will allow time for hydrocracking reactions, and so a minimum limit is nominally imposed on space velocity to avoid excessive yield loss. The net effect of decrease in feedrate is a reduction in both platformate and hydrogen yields (see table 3). In table 4, results of reactor outlet temperatures derived from Smiths kinetic model were further compared with the previous work by Lee et al. (1997) and Bommannan et al. (1989) related to reactor outlet temperatures predictions on the fixed-bed catalytic reforming processes. The temperature profiles prediction with Smiths model are more precise than the previous work with the maximum relative error within 0.9% as shown in table 4. The octane inferential model was developed based on the correlation derived from the model of the actual process. The expression was kept as simple as possible to enhance robustness and maintainability. When the model is nonlinear, the regression can still be carried out from the process data, because the equations are still linear with respect to the coefficients. The Research Octane Number (RON) is based on the following correlation derived from the actual model of the catalytic reforming process, retaining the process knowledge in the controller: RONmodel = [ PONA, Ti , F, RC , 1] + 2 (9)

4.00 10.0

3.80 9.60

5.3% 4.2%

TABLE 3. THE PREDICTION OF PRODUCT YIELDS AND REACTOR OUTLET TEMPERATURES FOR CASE 2: LOWER NAPHTHA FEED RATE Temperatures And Yields Outlet Temperature, Reactor (R1) Outlet Temperature, Reactor (R2) Outlet Temperature, Reactor (R3) Outlet Temperature, Reactor (R4) Yield of Reformate C5+, Vol % Production of Gases, Vol % Hydrogen LPG Model Results 448C 460C 487C 504C 80.60 Plant Test 447C 458C 485C 500C 83.20 Relative Error 0.2% 0.4%

In equation 9, PONA is the concentration of components in percent volume and Ti=1,2,3,4 are the four reactor inlet temperatures, F is the naphtha feed rate, Rc is the recycle gas feed rate, 1 is the
TABLE 4. COMPARISON OF REACTOR OUTLET TEMPERATURES (PREDICTED BY SMITHS MODEL) WITH OTHER PROCESS MODELS Reactor Outlet Temperatures Smiths Model Prediction Plant R1 (C) 450 446 441 440 434 434 R2 (C) 464 461 469 466 475 453 R3 (C) 485 486 479 480 489 476 R4 (C) 495 0.9% 493 505 1.2% Plant 499 4.9% Relative Error

0.4% 0.8% Lee et al. (1997) 3.2% Prediction

2.70 13.7

2.85 13.1

5.5% 4.6%

Bommannan et al. Prediction (1989) Plant

16

SAUDI ARAMCO JOURNAL OF TECHNOLOGY SPRING 1998

98.00 Prediction 97.00 Lab Results 96.00

95.00 Octane Nr.

94.00

93.00

92.00

91.00

90.00 1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 17

Time (hr) Fig. 2. Inferential model predicting reformate octane

filter of the first order to reduce the measurement noise, and 2 is a bias or trim factor to update the model from the laboratoryquality results. It is important to note that the equation 9 is nonlinear in terms of reactor inlet temperature dynamic response. The advantage of using a nonlinear model is that it is valid over a wider operating range and reduces the retuning effort. Equation 9 in the above form does not take into account pressure-temperature compensation. The temperatures in this type of correlation are normally pressure compensated; for that reason, the pressuretemperature compensation was carried out using the nonlinear Clausius Clapeyron equation given in Perry and Green (1984) and modified in the following form for this particular application: Tcomp = Tk Bi /(Tk LnP + Bi ) (10)

CONCLUSION
Smiths kinetic model has produced results which are significantly closer to the plant test results, as evident from table 1 to table 4. It was observed from the results of the model that was tested on the plant that low-pressure reforming has an increasing effect on the production of aromatics and also on the quantity of hydrogen. In addition, the plant test runs reveal that operation at low pressure increases the fraction of feed energy, in terms of gross heat of combustion, contained in the desired products, reformate and hydrogen. At higher feed rates (higher liquid hourly space velocity), fewer catalyst sites will be available per unit of feed, so higher reactor temperatures will be required for the same conversion (table 2). Finally, due to the difference in selectivity and activity of different catalysts and unit configuration, it is not possible to predict precise simulation of a specific unit by this method. The process model has also played a vital role in the control strategy. It was used to make the outputs follow their trajectories in multivariable control and to avoid the constraint violations. The octane inferential model was used to predict the reformate octane number, and it formed a closed-loop quality control, improving the dynamic performance of the system.

where Tcomp is a pressure-compensated temperature, Tk is the reactor temperature in Kelvin, P is the reactor system pressure, and Bi is a Clausius Clapeyron coefficient. Fig. 2 gives a comparison of octane inferential model prediction with the laboratory results. This inferential model was used in nonlinear multivariable control applications forming a closed-loop quality control which improved the dynamic performance of the system. The details of the nonlinear multivariable control and its application to catalytic reforming may be obtained from Ansari and Tad (1997).

SAUDI ARAMCO JOURNAL OF TECHNOLOGY SPRING 1998

REFERENCES
Ansari, R.M. and M.O. Tad. 1997. Constrained nonlinear multivariable control of a catalytic reforming process. Saudi Aramco Journal of Technology, Fall issue, pp. 34-42. Bommannan, D.; R.D. Srivastava; and D.N. Saraf. 1989. Modeling of catalytic naphtha reformers. Canadian J. of Chem. Eng. 67, p. 405. Lee, J.W.; K.Y. Ko; Y.K. Jung; and K.S. Lee. 1997. A modeling and simulation study on a naphtha reforming unit with a catalyst circulation and regeneration system. Computers and Chem. Engineering., v 21, suppl., pp. S1105-S1110. Perry, R.H. and D. Green eds. 1984. Chemical Engineers Handbook, Sixth Edition, 4-68. New York: McGraw Hill. Smith, J.M. 1970. Chemical Engineering Kinetics, 2d ed.; New York: McGraw Hill.

18

SAUDI ARAMCO JOURNAL OF TECHNOLOGY SPRING 1998