You are on page 1of 14

Control Design:

Fluidized Catalytic Cracking Units

Raphael Portela Chalhub December 18, 2013 CHBE 470

I.

Abstract

The report presented here attempts to better understand the control strategies employed in fluidized catalytic cracking units (FCCU) across industry. A process background is first introduced, followed by a general clarification of control strategy hierarchy employed for this equipment. Finally, studies are discussed and analyzed in order to better understand the effect of linear versus nonlinear control strategies in the multi-input and multi-output FCCU problem. In the study by Kalra and Georgakis, the simulation-based assessment of the linear MPC strategy had as motivation exploring whether linear strategies could be used in control design close to constraints where nonlinearities thrive and greatly affect closed-loop performance. It was found that although controls designed at operating points relatively close to the constraint can yield acceptable results (e.g. OP-II), there is no denying that as one approaches ever closer to the imposed restriction the effect of the disturbance becomes more pronounced and in some cases the variable never returns to its set point (e.g. OP-IV).Despite the findings in the first study, a interesting fact worth noting was that in order to optimize the economics of the FCCU process,
, it is many times useful to be able to operate as close to such restrictions as possible, a condition that is more easily achieved through nonlinear multivariable controls. Regardless, linear MPC strategies are widely used today. In order to better understand the differences between these two control strategies Study 2 was examined for it provided a good comparison of both. In the work by Ansari and Tade, it was found that indeed the nonlinear control outperformed the linear tests concerning decoupling capabilities and the behavior of the control during an inverse response. Although the nonlinear control behaved better, the reason why such controller is not more widely used is most likely due to its relative novelty and need for longer computing time or better computing capability.

II.

Introduction and Motivation

Fluid Catalytic Cracking (FCC) units are commonly used in the oil refinery industry to crack low value hydrocarbons into a range of more lucrative hydrocarbons. In fact, in U.S. refineries, the quantity of feed processed by FCC units is equal to 34% of the total crude oil treated in the United States. 1 Using a microspherical catalyst that behaves like liquid when properly fluidized, the unit is capable of converting high-boiling petroleum fractions known as gas oil into high-value fuels such as gasoline, jet fuel and diesel. Because of the importance of such unit in a refinery, it becomes essential to understand how process control instrumentation can be used to achieve a successful and efficient performance. The primary controls in the reactor-regenerator system, considered the heart of the equipment, are flow, temperature, pressure and catalyst level. Since differential pressure is used as the driving force to circulate the fluidized catalyst between the regenerator and reactor vessels, regulation of this variable is vital and can be achieved by utilizing a slide or butterfly valve in the regenerator flue gas line; the reactor pressure is controlled by the wet gas compressor. Another crucial control refers to maintaining a certain level of fresh catalyst, which can lose activity throughout the process; differential pressure

indicators are used to measure the catalysts raw levels, a reactor outlet temperature set point regulates the flow of clean catalyst and a slide or plug valve can extract spent reagent. Finally, the level of excess oxygen in the regenerator flue is monitored by manipulating the total air sent to the regenerator and the regenerator bed temperature is controlled by adjusting feed quality, preheat temperature, the use of recycle streams to the riser and the stripping steam rate assuming a complete combustion approach.2

III.

Background a. FCCU: Basic Process and Components


As mentioned earlier, the fluidized catalytic cracking unit has been one of the key processes petroleum refinery over the past couple of decades seeing as it is responsible for the majority of gasoline in an oil refinery.3 A simplified process schematic of such unit is displayed below:

Figure 1 Process Schematic of a Fluidized Catalytic Cracking Unit (FCCU) As shown, feed oil from the crude unit goes through a heat exchanger and is sent to a riser tube where it makes contact with the hot regenerated catalyst. This interaction causes the heavy oil feedstock to crack into lighter hydrocarbons, which in turn vaporize and are then stripped from the catalyst through the aid of a steam supply. The spent catalyst, which has lost its activity, must be sent to the regenerator for removal of carbon in order to be recycled back into the reactor.4

The two main components of the FCC system are the reactor and regenerator, which interact in a circuit-like manner. Nevertheless, other major components exist and are fundamental to understanding the FCCs control system. These are listed below together with a brief description of their functions: Reactor-Riser: allows for catalyst to mix with oil resulting in extraction of lighter hydrocarbons Regenerator: allow for catalyst to regain its activity by removing carbon or inserting new catalyst Regenerator Catalyst Hopper: allows for catalyst to be aerated before entering standpipe; the catalyst can then achieve its maximum flowing density and consequently greater pressure buildup at standpipe Regenerated and Spent Catalyst Standpipe: provides the pressure to drive the catalyst from regenerator to reactor and from reactor to regenerator respectively Regenerated and Spent Catalyst Slide (or Plug) Valve: regulates flow of catalyst to riser and to regenerator respectively; maintains pressure head in standpipe and impedes flow reversal5

b.

Fundamentals of FCC Process Control and Instrumentation


The operation of the FCC is highly dependent on pressure seeing as the movement of catalyst throughout the system occurs due to pressure differentials. The slide or plug valve found in the regenerator flue gas line is used to control the differential pressure between the regenerator and the reactor. Furthermore, the pressure in the reactor is regulated by the wet gas compressor (WGC).5 In addition to circulation, the catalyst must also be replaced due to losses at the vessels. While the catalyst level in the reactor is controlled by the two valves, excess catalyst is extracted when fresh amounts are introduced through a controlled periodic withdrawal of such. Differential pressure indicators are placed in both main vessels to measure such levels. For the introduction of clean catalyst, regulation is done via temperature measurements of the reactor and riser. 5 In order to understand the complex process control strategies discussed later, it is essential to be familiar with the general control hierarchy employed in the FCC system: basic and advanced level control systems. The basic level, also known as the regulatory control system, has as its main purpose to ensure the safe operation of the system. These process controls are usually simple linear control loops with proportional-integral-derivative (PID) actions determined by higher levels in the control hierarchy. The advanced or supervisory level systems serve the purpose of optimizing the process by coordinating various loops and working with a significant number of degrees of freedom. Finally, the highest level pertains to the plant wide optimization; this optimization is run at steady state and performed off-line at regular intervals. A diagram illustrating such relationships is shown below:6

Figure 2 Hierarchy Scheme of FCCUs Control Strategy The primary controls of the regulatory level are flow, temperature, pressure and catalyst level. Stream flows are monitored by flow controllers, which set desired flows for fresh feed, recycle, air rate, stripping steam, etc. by regulating valves. The temperature in the reactor is monitored by a temperature controller that regulates the regenerated catalyst slide (or plug) valve. The temperature in the regenerator depends on the catalyst regeneration mode: partial or total combustion. For partial combustion, the temperature is monitored by adjusting the flow of air into the vessel. For total combustion, the temperature varies according to a series of specifications such as feedstock quality, catalyst properties, use of recycle, stripping steam rate and others. Reactor pressure is indirectly controlled by the pressure control on overhead receiver responsible for controlling the wet gas compressor. On the other hand, the regenerator pressure is regulated by a pressure controller responsible for adjusting the flue gas slide (or butterfly valve). Next, the catalyst level in the reactor is regulated by a level controller responsible for tuning the spent catalyst slide valve, while the catalyst level in the regenerator is manually altered in order to maintain catalyst inventory. Finally, it is important to note that sometimes a set point for spent or regenerated catalyst may not reached despite complete opening and closing of their particular valves. Because such extreme action could result in an unstable pressure, the basic control will also encompass a low differential pressure override for safety. This avoids phenomena such as flow reversal which could result in undesirable equipment damage.5 The supervisory control level allows for maximization of profit by running the unit concurrently against many constraints. These variables include factors such as limits on air blower, wet gas compressor, slide valve differentials and many others. In the case of the FCC unit, this usually involves

a model-based system that uses a multivariable process to obtain the optimal operating conditions. Although such level involves much more intricate control algorithms, it is at such higher level where most of economic benefits can be obtained. Nevertheless, the lower level restrictions must always be met to ensure the proper functioning of the control.6 Due to the relative simplicity of the basic control level, most of the recent research and development pertaining to the process control of the FCC unit focuses on improving the existent advanced control schemes because not only does it pose more of a challenge, but also yields the greatest monetary return. The following section attempts to compare a series of different high level control strategies providing insights into their advantages and disadvantages.6

IV.

Experimental Work
With respect to the advanced control level for the FCC unit, perhaps the most widely studied and wellaccepted strategy is the model predictive control (MPC). The appeal of adopting a MPC stems from its intrinsic anticipative capability and its ability to explicitly handle constraints in controlled and manipulated variables, two factors that ensure better control performance relative to the decentralized PID control. a. Fundamentals of Model Predictive Control In a basic process control course, most models studied are based on continuous time. In reality, however, manipulated input changes are made at discrete time intervals and measured outputs are available at discrete sample times. MPC is defined as a discrete autoregressive model since it assumes that the output at a specific time step is a function of the outputs and inputs at previous time steps.4 The basic idea behind the MPC is the necessity to control a multi-input and multi-output process while avoiding violations of the inputs and outputs constraints. Furthermore, the strategy prioritizes some output variables and ensures that these reach their optimum while keeping other variables of less economic significance within their respective limits. Taking advantage of dynamic models in combination with measurements when available, the MPC is able to accurately predict with the output variables, also known as controlled variables (CVs). From such information, suitable changes in the input variables can be made to ensure that the desired conditions are met. Input variables are many times referred to as manipulated variables (MVs). A block diagram for the model predicted control is shown below:7

Figure 3 Block Diagram for Model Predictive Control The main advantages of adopting a MPC include: (1) model can accurately predict dynamic and static interactions between all types of variables, (2) variables constrains are met, (3) control calculations are coordinated together with calculation of optimum set points and (4) model predictions can anticipate potential problems allowing for preventive measures to be taken.7 One of the main requirements for the use of an MPC is the presence of a model. These can be broken down into two main categories: linear (LMPC) or nonlinear (NMPC). Although LMPC tends to be more popular due to its relative simplicity and reduced computational effort, nonlinear MPC can provide significantly better control especially for processes that exhibit substantial fluctuations.8 The following sections b and c discuss different studies done regarding MPC strategies of FCC units. The first study assess the effect of utilizing a linear MPC given a nonlinear process. The second study evaluates the effectiveness of a nonlinear MPC despite its greater intricacy. b. Study 1: Performance of Linear MPC given Process Nonlinearity The study performed by Kalra and Georgakis analyzes the possible restrictions imposed by process nonlinearities with regard to a linear MPC strategy. The algorithm was tested in a Model IV FCC unit and the model used to capture the process nonlinearity was developed through a cooperative effort between Lehigh University and the Amoco Corporation enabling the capture of major dynamics of the unit while taking into account equipment and operating constraints due to economic, environmental and safety factors. Since the ability to reject disturbances, especially at high throughputs, is a major goal of the FCCU control thus avoiding violation of restrictions, tests of variable responses to disturbances were performed. Four different operating points (OP) corresponding to four different throughputs were used.9 After careful consideration, four CVs and four MVs were selected for the multiple-input and multipleoutput (MIMO) predictive control strategy. The four manipulated variables and reasoning behind their selection are summarized below: 1. lift air flow rate (F9set): has immediate effect on oxygen available for burning coke off the catalyst

2. feed flow rate (F3set): enables the operation to remain at total combustion 3. slurry recycle flow rate (F4set): enables the operation to remain at total combustion 4. reactor-regenerator differential pressure (Dpst): allows for change in catalyst recirculation9 The four CVs are listed below together with the rationale behind their selection: 1. stack gas carbon monoxide concentration (COsg): environmental constraint imposed on control design; regulation require that it not exceed 300 ppm 2. reactor riser temperature (Tr): affects the yield of wet gases produced 3. regenerator bed temperature (Treg): best controlled by F3set and F4set, it is desirable to keep such variable above 1265 oF for total combustion operation 4. wet gas compressor suction valve position (V11): it allows for more air to be pumped into the system in order to increase the FCCUs throughput9 In addition to following the general MPC algorithm, a tuning of the controller was also performed. Parameters included number of time steps into the future over which error from set point has to be minimized (prediction horizon), number of control moves into the future calculated by the controller (controller horizon), and others. Furthermore, the use of matrices was employed in order to adjust for factors such as relative importance of tightness of control on the outputs as well as magnitude of control moves.9 c. Study 2: Comparison between Nonlinear Multivariable Control and Linear MPC As shown in Study 1, the linear MPC performed relatively well given that it did not operate too close to the imposed constraint. In order to optimize the economics of the process, however, it is many times useful to be able to operate as close to such restrictions as possible, a condition that is more easily achieved through nonlinear multivariable controls. Regardless of such flaw, the fact remains that linear MPC strategies are widely used today and thus it is important to compare these two methods in order to verify if constructing a more intricate control is indeed advantageous despite its own drawbacks. Similar to the linear MPC, the nonlinear multivariable control (NMC) must solve challenging tasks: (1) account for the multivariable character of the process interactions, (2) predict the nonlinear behavior of the process and (3) run within the material and general operating constraints.10 Such condition is achieved by manipulating the following variables: combustion air flow, feed flow rate, feed preheat temperature and riser outlet temperature. These MVs are used to control both the flue gas oxygen concentration and the regenerator bed temperature. Underlying this optimization are imposed constraints regarding flue gas flow, riser outlet temperature and wet gas compressor suction pressure as well as disturbances in the feed composition. A schematic of the NMC problem is shown below:10

Figure 4 Schematic of the NMC Problem for the FCC Reactor-Regenerator System The nonlinear model explored in this study takes advantage of a Generic Model Control (GMC), which employs a nonlinear process model directly within the controller. The Optimization toolbox in MATLAB was used in order to optimize the nonlinear functions by finding the constrained minimum of the functions of several variables. In addition to that, further optimization was performed by updating model parameters in order to reduce the mismatch found between the model and the true process.10 The linear multivariable control compared to in this study was designed by using the MATLAB toolbox on MPC for linear control application combined with some control techniques from algorithms such as Dynamic Matrix Control (DMC), a first generation of MPC systems.7, 10

V.

Results and Discussion a.


Study 1 Results and Analysis For the MIMO control problem presented, a linear MPC was used due to its simplicity and ability to handle multivariable process as well as constraints. Four different OPs numbered I through IV corresponding to four different throughputs were used.9 Although the report performs a series of different tests including the verification of the process nonlinearity, evaluation of altering tuning parameters and a couple of others, the report presented here focuses on assessing the effectiveness of the designed linear MPC after optimal parameters have been defined.

In order to gauge the LMPCs performance, a dynamic model was created at each OP and then the MPC algorithm was employed in the controller design. The response plots to a feed coking factor disturbance for the four CVs (on the left) and four MVs (on the right) are shown below:9

Figure 5 MPC responses at OP-II (dash-dotted), OP-III (dashed), and OP-IV (solid) Aside from the spike in the COsg plot immediately after the disturbance is introduced, the MPC performed fairly well at OP-II (feed flow rate = 127.3 lb/s) seeing as all CVs return to their starting values after at most 120 minutes. Recall that the goal is that the control ignore such disturbances.9 However, as the throughput increases, the ability of the MPC to reject the same disturbances worsens. This can be noticed by examining the responses for OP-III (feed flow rate = 128.9 lb/s) and OP-IV (feed flow rate = 131.0 lb/s). The oscillatory behavior is perhaps most pronounced in the OPIV responses.9 Thus, the tests performed show that a LMPC performs well as long as the conditions are not too close to the constraints. Operating the process closer to its restrictions yields oscillatory behavior and in some instances the variables never return to their starting values (e.g. Tr at OP-IV).9

b.

Study 2 The simulation-based tests yielded a series of results for both the nonlinear and linear control algorithms. The first set compares both controls relative to their decoupling potential, or the ability to eliminate control loop interactions. It is shown below:10

Figure 6 Comparison of Decoupling Capability of the Linear and Nonlinear Controllers As can be seen from the plots, the nonlinear control reaches the steady state quicker in both controlled and manipulated variable responses. This is due to the fact that FCC process is nonlinear and time-variant, thus being more easily captured by the nonlinear algorithm. Another test performed pertains to controls performance of both algorithms in dealing with an inverse response. The results are summarized graphically below:

Figure 7 Comparison of linear and nonlinear algorithms in controlling the process with inverse response

Note that again, the performed test shows how the nonlinear strategy tends to behave more advantageously reaching the desired steady state value in less time yielding less error. In the case of a inverse response, this can be rationalized by noting the presence of right-half place (RHP) zeroes in the linear control transfer functions, which in turn become poles of the process inverse, a scenario that can pose some difficulty. VI. Summary and Conclusions In the study by Kalra and Georgakis, the simulation-based assessment of the linear MPC strategy had as motivation exploring whether linear strategies could be used in control design close to constraints where nonlinearities thrive and greatly affect closed-loop performance. It was found that although controls designed at operating points relatively close to the constraint can yield acceptable results (e.g. OP-II), there is no denying that as one approaches ever closer to the imposed restriction the effect of the disturbance becomes more pronounced and in some cases the variable never returns to its set point (e.g. OP-IV).

Despite the findings in the first study, a interesting fact worth noting was that in order to optimize the
economics of the FCCU process, , it is many times useful to be able to operate as close to such restrictions as possible, a condition that is more easily achieved through nonlinear multivariable controls. Regardless, linear MPC strategies are widely used today. In order to better understand the differences between these two control strategies Study 2 was examined for it provided a good comparison of both. In the work by Ansari and Tade, it was found that indeed the nonlinear control outperformed the linear tests concerning decoupling capabilities and the behavior of the control during an inverse response. Although the nonlinear control behaved better, the reason why such controller is not more widely used is most likely due to its relative novelty and need for longer computing time or better computing capability.

VII.

Suggestions for Future Work A series of future work approaches can be explored regarding the topic of control design in FCCU units. First of all, there are a myriad of different studies being done on various nonlinear control systems for FCCU units. For instance, it would be wise to explore the work done by Cristea et al. in terms of selecting the best controlled and manipulated variables for developing a nonlinear MPC. In the paper titled Simulation and model predictive control of a UOP fluid catalytic cracking unit, eleven different control schemes are tested ranging from 3 MVs x 3 CVs all the way to 6 MVs x 5 CVs. One of the best resource for a compilation of all sorts of studies done on control design FCCU can be found in Pinheiro and Fernandes work shown in the reference section at the end of the report.

VIII. Symbols and Abbreviations


CV = controlled variable Fa = regenerator air rate [in Fig. 1] FCC = fluidized catalytic cracking FCCU = fluidized catalytic cracking unit Fs = catalyst recirculation rate [in Fig. 1] MIMO = multiple-input and multiple-output MPC = model predictive control MV = manipulated variable NMC = nonlinear multivariable control

OP = operating point Tcy = regenerator gas temperature [in Fig. 1] Tcysp = regenerator gas temperature set point [in Fig. 1] T1 = reactor (separator) temperature [in Fig. 1] T1sp = reactor (separator) temperature set point [in Fig. 1]

IX.
1D.

References

Nakimura, Oil Gas J., 62 (Dec. 23, 2002). G., P. Indumathi, and V. Selvakumar. "Design of Controllers for a Fluidized Catalytic Cracking Process." Chemical Engineering Research and Design. 88. (2010): 875-880. Web. 18 Dec. 2013. 4Bequette, B. Wayne. Process Control: Modeling Design, and Simulation. 1st ed. Saddle River: PrenticeHall, 2002. Print. 5Sadeghbeigi, Reza. Fluid Catalytic Cracking Handbook: An Expert Guide to the Practical Operation, Design, and Optimization of FCC Units. 3rd ed. Oxford: Elsevier, 2012. Print. 6Hovd, M., and S. Skogestad. "Procedure for Regulatory Control Structure Selection with Application to the FCC Process." AIChe Journal. 39.12 (1993): 1938-1953. Web. 18 Dec. 2013. 7Seborg, Dale E., Thomas F. Edgar, and Duncan A. Mellichamp. Process Dynamics and Control. 2nd ed. Hoboken: John Wiley & Sons, 2004. 534-566. Print. 8Pinheiro, Carla I.C., Joana L. Fernandes, et al. "Fluid Catalytic Cracking (FCC) Process Modeling, Simulation, and Control." Industrial & Engineering Chemistry Research. 51. (2012): 1-29. Web. 18 Dec. 2013. 9Kalra, Lokesh, and Christos Georgakis. "Effect of Process Nonlinearity on the Performance of Linear Model Predictive Controllers for the Environmentally Safe Operation of a Fluid Catalytic Cracking Unit." Industrial & Engineering Chemistry Research. 33. (1994): 3063-3069. Web. 18 Dec. 2013. 10Ansari, R.M., and M.O. Tade. "Constrained nonlinear multivariable control of a fluid catalytic cracking process." Journal of Process Control. 10. (2000): 539-555. Web. 18 Dec. 2013.
3Pandimadevi,

You might also like