FCCU ADVANCED CONTROLS INCREASE FEED RATE AND COLUMN STABILITY

Gareth Rowlands
Conoco U.K. Ltd.
Humberside, England

Aydin Konuk, Frank Kleinschrodt
Setpoint Inc.
Houston

The application of advanced process controls to the fluid catalytic cracking unit (FCCU) at Conoco U.K. Ltd.'s Humber refinery boosted the unit feed rate, reduced flooding, and stabilized the column.

This article reports on the system's performance over a full year of operation on the FCCU's vapor recovery unit, main fractionator, and reactor/regenerator.

BACKGROUND

Conoco U.K. Ltd. initiated a study in 1987 to determine possible control strategies, estimate their benefits, and develop project costs. The results indicated rapid payback from the implementation of comprehensive control of the refinery's FCCU.

After funding was approved, Setpoint Inc. began designing the recommended controls in late 1988. It completed the design in the spring of 1989. After a cycle of careful review, comment, and approval by Conoco, the programming was initiated. It was completed in late summer 1989.

Following a factory demonstration and functional performance test, commissioning of the controls began in December 1989. The bulk of the commissioning was completed in March 1990.

The Humber FCC project was characterized by significant client involvement in every phase.

At the study phase, it is essential to capture the experience of the client's engineering and operations staffs when establishing control objectives and strategies. Engineers know why variables affect one another, while operations people know what works and how to operate the unit.

Conoco used a combination of senior operations people and experienced process engineers to fill this need. At the detailed design review, all strategy details were again reviewed and improved by the same persons who had contributed during earlier stages of the study.

At the demonstration of the controls, the same people again ensured that the controls responded as predicted and that the operator interface with the controls was understandable and efficient. After completing the project, the Conoco team had the skills to tune and modify the controls to suit the varying requirements of the unit.

Anyone familiar with FCCUs knows how much flexibility is inherent in an FCCU and how it is called upon to adjust to the needs of the refinery. An FCC control system must be built with enough flexibility to adjust to a wide range of anticipated conditions. It must also be reconfigurable to adapt to conditions not anticipated.

Furthermore, the refinery must have staff on site or readily available to keep the system in good working order. This is a big commitment for a refiner, and it is absolutely essential to the longterm success of an advanced controls project.

A senior operator was assigned to the commissioning team. The role played by this man was crucial to the speed with which the controls were applied and accepted by the operators.

Two Conoco process engineers and one Conoco systems specialist were also available essentially full-time to support the commissioning effort and to become familiar with Setpoint's commissioning techniques. While this might seem like expensive overkill to refiners who run with extremely lean staffs, the benefits of the investment clearly justified the effort.

As an added benefit, there is now a base of personnel within Conoco that can train others. This enables the company to proceed into the advanced control arena more quickly, while taking care of the previously commissioned controls. Within a year of completing commissioning of the FCCU project, Conoco initiated four multiunit studies and three multiunit projects.

VRU CONTROL STRATEGIES

An overview of control strategies implemented on the vapor recovery unit (VRU) is shown in Fig. 1. Because many of the strategies are similar, they will be discussed in strategy groups rather than by individual applications.

NONLINEAR LEVEL CONTROLS

Distillation columns run much more smoothly when feed rates are constant or change slowly. The objective of nonlinear level controls is to use the volume in the various VRU vessels to absorb disturbances from the upstream part of the unit, and to change flows from surge vessels only when necessary.

Setpoint's proprietary control algorithm provides the following level control characteristics:

When the level is near the setpoint, very small changes are made to the flow controller target.
When the level is at predefined span limits, aggressive control action is taken to avoid overfilling or drying of the vessel.
When the level is between its setpoint and span limits, the aggressiveness is increased as it moves toward the span limit.

Tuning parameters for the nonlinear algorithm are easy to understand. They include the desired size of the disturbance to be handled and a process gain that can be either analytically derived or developed from a simple step test.

CONSTRAINT CONTROLS

Minimization and/or maximization constraint controllers come in two basic flavors: those that sit under advanced regulatory controllers to protect against occasional constraints, and those that sit at the top of an advanced control strategy and push the unit to a limit.

Most of the constraint controllers in the VRU do the latter. They are intended to push a variable, such as column pressure, to a limiting value. The direction in which they push is usually the result of a presolved optimization problem, for which the desired result can never be achieved because of a unit limitation. These limitations are included as constraints in the constraint controller.

Setpoint's proprietary constraint controller for the Honeywell TDC-3000 system handles both basic modes of constraint controllers. It can handle two-sided constraints as well as resolve conflicts between constraints of differing levels of importance. It can also handle constraints with complicated dynamics, including dead times.

An example of a maximum constraint controller is the maximum-pressure constraint controller on the secondary absorber. It is intended to maximize the recovery of C3 from the fuel gas, subject to stripping constraints at higher pressures.

An example of a minimum constraint controller is the stabilizer minimum-pressure constraint controller. It is intended to reduce the heat-input requirement of the stabilizer (for main fractionator considerations), and to improve the separation in the column, subject to column limits like flooding and rundown valve limits.

COMPOSITION CONTROLS

Model-based composition controls are used at every location where a final key product specification is measured with an on stream analyzer. A dynamic model is developed between a continuously measured value, like column temperature, and a relatively infrequently measured value, like a gas chromatograph reading. The model is updated at each analyzer result.

The model runs on a scheduled, frequent basis and calculates a temperature that will produce the desired composition. This approach handles infrequent analyzer results as well as significant dead time between the column temperature change and its effect on the analyzer result.

Most of the model-based control strategies were easy to implement once the analyzers were proved to be reliable. An exceptionally difficult one was on the overhead of the C3/C4 splitter.

This column is subject to significant disturbances when an intermittent regeneration stream from the alkylation unit is routinely routed into its feed drum-the stabilizer overhead accumulator. A revised scheme is planned where a feed composition analyzer will provide feed-forward control.

ANTIRECYCLE CONTROLS

Many VRU designs tend to move into a recycle condition in the stripper/absorber area. As part of the Humber VRU controls, quality objectives are temporarily sacrificed to ward off an incipient recycle condition. When the recycle problem is resolved, the unit is redirected back to the quality objectives.

This is an important advanced regulatory control that substantially increases the throughput flexibility of an FCC VRU. The override function is only active during short-term transient events unless unreasonable ethane rejection and C3 recovery targets are simultaneously pursued.

The recycle problem was a common condition in the unit prior to implementing this function. Its frequency has been greatly reduced, even at the higher feed rates now employed.

FCC MAIN FRACTIONATOR

An overview of the FCC main fractionator control strategies is shown in Fig. 2. The strategies include:

Gasoline/light cycle oil (LCO) cutpoint control
LCO pan level control
Steam/LCO ratio control
Shed vapor-temperature control
Slurry quench constraint control.

The objectives of the controls are to assist the operators in meeting product quality specifications (gasoline and LCO endpoints) and to maximize the yield of the more-valuable products, within column constraints.

This is achieved by first stabilizing the column operation using the LCO pan level control, steam/LCO ratio control, shed vapor-temperature control, and slurry quench constraint control. The product cutpoint controls then make slow adjustments to column temperature-profile variables to keep the products on specification.

Operators enter specification targets. These targets do not normally require adjustments for feed rate and/or feed quality changes or for column pressure changes.

The gasoline/LCO cutpoint-control strategy uses a calculated (inferred) gasoline endpoint to manipulate the heat balance at the LCO pumparound, which in turn affects the material balance between gasoline and LCO. The inferred variable takes into account varying yields of affected products, column pressure, partial pressure variation, and internal traffic (separation effects).

The manipulated variable (MV) is LCO pumparound flow. This is a slightly unusual choice of MV, but tests during commissioning showed that it provided a more-stable column operation than the typically used top temperature.

To make this strategy work, the top temperature is controlled manually so that only one effective temperature controller is active at the top of the column. The operators find that, to obtain significant changes in column operation, they must occasionally adjust the top reflux flow to keep the LCO pumparound flow in an acceptable working region.

The function was not automated because of operator acceptance of this strategy and lack of direct economic benefits. Because the performance of this loop achieves product specifications close to laboratory tolerances, the operators use statistical process control charts to make occasional changes in the loop target only when there is a statistically significant deviation from product specifications.

Fig. 3 shows the trend of a recently updated gasoline/LCO cutpoint correlation. The data were taken in a gasoline-minimization mode, where the cutpoint was varied over a fairly wide range.

The performance of this control is expected to have a standard deviation of approximately 2.5 C. Conoco's Humber laboratory has a standard deviation of about 1 C. on gasoline samples. And the IP limit for repeatability is 4 C.

It is not likely that an inferential control system at the Humber site could achieve a standard deviation much better than 2 C. The only possible way to improve current performance would be to use a continuous analyzer with precision comparable to Humber's laboratory. Currently, there is no recommendation to do this because of the purchase and maintenance costs of an endpoint analyzer.

For the steam/LCO ratio-control strategy, the stripping steam rate is adjusted to maintain a constant ratio between it and LCO product flow. This prevents waste of steam (overstripping) at reduced unit flow rates and keeps the stripping steam at a predetermined optimum level.

It is well known that stabilization of the slurry heat removal, and consequently the shed-tray temperature, help stabilize the entire back end of an FCCU. Vapor-temperature cutpoint control achieves that goal and much more.

The lowest level of this strategy is a ratio controller, which maintains a constant ratio of the slurry-pumparound flow to a dynamically compensated unit feed rate.

The target for the ratio controller is set by a shed tray vapor-temperature controller.

Experience with this controller showed that the lower-column temperature profile could be stabilized to a standard deviation of less than 10 F. The target for the vapor-temperature controller is in turn set by a constraint controller that attempts to satisfy the LCO/slurry cutpoint controller, subject to the slurry quench valve being set in an operable region.

During the early days of this strategy's operation, the LCO/slurry cutpoint was maintained in a narrow region. However, at the increased unit throughput seen with the total advanced control package, the LCO rundown system could not handle the new LCO rates. For this reason the LCO endpoint actually runs below its maximum specification.

The objective of the slurry quench constraint-control strategy is to minimize the slurry quench used, subject to a maximum bottoms temperature and minimum total slurry flow on the main fractionator. This makes the maximum amount of slurry available for direct heat exchange against the reactor effluent. This in turn helps achieve maximum unit capacity when the unit is limited by main fractionator heat removal.

REACTOR/REGENERATOR CONTROL

An overview of the FCCU reactor/regenerator control strategies is shown schematically in Fig. 4. The strategies include the following:

Stripping steam/catalyst circulation-rate ratio control
Pressure-balance control
Combustion/severity control
Unit optimization.

The steam-to-catalyst ratio control strategy adjusts the stripping steam rate to maintain a constant ratio of steam to a calculated catalyst-circulation rate. This ensures adequate stripping of hydrocarbons without unnecessarily loading the main fractionator or wasting steam through overstripping.

The pressure-balance controller is a two-dimensional constraint controller customized for FCC use. It adjusts the reactor/regenerator differential-pressure controller and the main fractionator overhead-receiver pressure controller setpoints.

The controller resides in the process control computer and runs once every 5 min. It is designed to minimize both the regenerator and reactor pressures, subject to unit constraints.

Pressure-balance control calculations are based on steady-state models obtained from plant tests conducted during commissioning. The pressure-balance controller is tuned to move the unit pressures gently.

Combustion/severity controls are implemented using a single multivariable controller, Idcom-M.

Idcom-M is a generic model-based, predictive, multivariable, multiobjective control algorithm that allows the user great flexibility in designing a strategy to meet a number of objectives.

The first objective is to maintain MVs within position and velocity limits, followed by maintaining controlled variables (CVs) at setpoints and/or within limits. The last objective is to maintain MVs at setpoints.

Idcom-M resides in the process control computer and runs once per minute. The control calculations are based on dynamic models between CVs and MVs obtained during commissioning by a process identification package.

Idcom-M is tuned to move its MVs quickly in response to disturbances, and slowly to approach MV setpoints when constraints allow. Idcom-M MVs include regenerator air rate, fresh-feed rate, and reactor temperature.

Pressure balance and Idcom-M controllers, although implemented as stand-alone software packages, are designed and tuned to work together to control the reactor /regenerator. The controlled variables included in these two strategies are listed in Table 1. Most are included in both pressure balance and Idcom-M.

Table 1 also lists the pressure balance and Idcom-M manipulated variables. Controlled variables in this application were all assigned maximum, or maximum and minimum limits.

The FCC unit optimizer is based on a nonlinear, steady-state FCCU model. The model is initially tuned off-line using FCCU equipment and operating data. The optimizer runs once every 6 hr. At the beginning of each run, key model parameters are updated using current 2-hr-average process data.

The optimizer maximizes FCCU gross margin subject to constraints, which include most of the reactor/regenerator constraints shown in Table 1, with the following additions: gasoline octane (RON), propane/propylene production, butane/butylene production, gasoline production, LCO production, slurry oil production, and fuel gas rate.

The optimizer resides in the process control computer. The optimizer independent variables are feed rate, reactor temperature, catalyst-cooler duty, and feed preheat.

The optimum feed rate and reactor temperature calculated by the optimizer are passed to Idcom-M as MV setpoints. The optimum catalyst-cooler duty is implemented by a catalyst-cooler duty constraint controller.

The optimum feed preheat is implemented by the operator. Because of reactor stripper capacity limitations, hydrocarbons on spent catalyst cannot be adequately removed at low feed-preheat values, making the use of preheat as an optimization variable impractical.

Reactor/regenerator delta-P and fractionator overhead pressures are not included in the unit optimization as independent variables because the solution is known a-priori. Minimizing these two variables maximizes profit (lower blower power cost and better yields benefits exceed higher wet-gas compressor costs).

DCS/COMPUTER ENVIRONMENT

The primary control platform for this project was a Honeywell TDC-3000 applications module (AM). Access to all control functions is provided by the Honeywell Universal Station. The operators of the FCCU have access to four Honeywell screens for regulatory and advanced control activities.

Approximately 100 custom touch-screen graphics displays were developed for the project. Conoco enhanced its process overview graphics to include a summary block that indicates the status of each strategy.

Color, reverse video, and blinking features were used to give the operator an accurate bird's-eye view of all of the advanced control strategies. From the overview, graphics touch points were assigned to allow the operator to quickly access more detailed screens.

A secondary platform for the project was a Digital Equipment Corp. VAX 3800 minicomputer running the Setcon database and control package. The multivariable control and unit optimization functions were resident on this platform. Setpoint provided the interface between the computer and the DCS computer gateway.

All on/off functionality, as well as target-setting and major tuning parameters for the computer-resident functions, were provided in the DCS. This is called a "single operator window" concept. Additional engineering access was provided by a PC-based computer console system, as well as standard VAX terminals.

RESULTS AND BENEFITS

The commissioning of the advanced controls also coincided with a change of unit catalyst and an improvement in feed availability and quality. All three changes helped to boost unit feed rate and reactor severity. It is estimated that the long-term benefit of advanced control has been worth 3,000 b/d of extra feed rate since February 1990.

In addition, during the development of the advanced control strategies, certain areas of current operation were highlighted for improvement. Operator intervention into the main fractionator and gas-plant control loops has been reduced, leading to steadier units and product quality improvements.

OPERATOR ACCEPTANCE

The FCCU operators had been skeptical about whether the controls would work, and believed that frequent intervention would still be needed.

The first loops to be commissioned were the nonlinear level controllers on the VRU surge drums and tower bottoms. The panel operators had often spent time "chasing" these levels, especially during flow oscillations after feed-rate changes.

The nonlinear level controllers stabilized the unit considerably, and instances of empty or overfull vessels were reduced. This was the first important step in demonstrating to plant personnel what the advanced controls could do for their plant.

Although the quality improvements achieved with the applications on the main fractionator and VRU were seen as important, the commissioning of the multivariable controllers on the reactor /regenerator had the most impact on the panel operators. Within a week, all had confidence in the controller's ability to move feed rate, riser and/or unit pressures to stay within unit constraints.

Time spent monitoring these constraints was reduced on each shift. More importantly, the factor of safety-operating margins to prevent unsuspected excursions beyond these limits was reduced substantially or eliminated.

RUN TIME

Operator confidence is reflected in the run-time figures for the various strategies. The multivariable controllers on the reactor/regenerator have on stream factors of more than 98% and have only been taken off-line during times of instrument calibration or feed shortage.

Conoco has yet to record an instance of either computer hardware or software failure.

Time on-line for the main fractionator and VRU loops is also high, although some time has been spent tuning the cutpoint calculations on the main fractionator to produce reliable results under all operating scenarios.

The stabilizer controls spent some time off-line last winter when economics dictated minimization of alkylation unit feed.

VRU CONTROLS

VRU controls allowed stabilization of the unit and maintenance of product qualities in the presence of higher and varying feed rates. Benefits have been quantified only for the stabilizer overhead C5 controls, which resulted in approximately 55 b/d of additional LPG.

MAIN FRACTIONATOR CONTROLS

The primary benefit of the main fractionator controls resulted from the stabilization of the column, which allowed much higher throughput. This column had a tendency to flood at unit rates of approximately 40,000 b/d prior to applying advanced controls.

This was displayed by an increase in pressure drop in the column and a corresponding increase in regenerator pressure and loss of air capacity. The column has now achieved rates as high as 48,000 b/d.

Flooding is an asymmetric, dynamic issue. One quick move or disturbance can move a column into flooding in minutes, and it can take hours to stop it without a significant reduction in feed rate. Column stabilization means that the real steady-state flooding conditions can be much more closely approached.

Two major factors reduced flooding. The first is that, at stable shed-tray temperatures, the entire column and gas plant run more smoothly. Interaction with the VRU controls significantly affects column flooding, and much time had previously been spent tuning the VRU controls to minimize the impact on the fractionator, and vice versa.

The second factor is the change in the LCO draw system.

Removal of liquid disturbances in the column by drawing them out as LCO product has little impact on quality variation, but has a significant impact on reducing the variation in net reflux in the section of the column between the LCO draw and the HCO pumparound.

The secondary benefit of the main fractionator controls is that the inferential controls allowed the reactor/ regenerator controls to make significant variations in feed rate and column pressure, with no degradation in observed product quality.

In fact, the average standard deviation of the gasoline endpoint (target minus lab) has been observed to have a 2-month average value of 2.68 C., as compared to a precontrol deviation of 3.87 C. for a comparable time period.

The estimated yield improvement of gasoline for this change in standard deviation is about 0.35 vol %, or 160 b/d. No benefit is being claimed at this time for LCO yield improvement because the rundown system limits LCO production at the higher unit throughput.

REACTOR/REGENERATOR CONTROLS

Although pressure balance and Idcom-M are multivariable controllers, they were designed to act like five one-dimensional constraint controllers at steady state. In other words, each MV controls a group of constraints at steady state.

The following discussion will help interpret the longterm performance of the reactor/regenerator controls.

Air rate controls the following blower/regenerator-related constraints: blower speed, turbine governor position, expander bypass-valve position, flue gas oxygen, and cyclone velocity, The most common limiting constraints are cyclone velocity and turbine speed.

Riser temperature controls these wet gas compressor-related constraints: wet-gas compressor power and wet-gas compressor inlet throttle-valve position.

Throttle-valve position is the more active constraint, and compressor power is very near its limit.

Feed rate controls these main fractionator and regenerator-related constraints: main fractionator delta-P, main fractionator Slurry-pump-around valve position, and regenerator dense-bed temperature. In this group, main fractionator delta-P (flood) is the most active constraint. Slurry-pumparound valve position (bottoms-cooling constraint) has also occasionally been a limiting constraint.

Main fractionator overhead pressure controls these regenerator pressure constraints: regenerator pressure and blower discharge pressure. Blower discharge pressure is the limiting constraint.

Reactor/regenerator pressure controls these catalyst circulation-related constraints: regenerated catalyst slide valve delta-P, regenerated catalyst slide valve position, spent catalyst slide valve delta-P, and spent catalyst slide valve position. The limiting constraints among these are regenerated catalyst slide valve delta-P and valve position.

Both pressure balance and Idcom-M are multivariable controllers, and both will move multiple MVs for dynamic control purposes. For example, if a disturbance causes a wet gas compressor-inlet throttle valve to exceed its maximum limit, then pressure balance will increase the fractionator-overhead pressure while Idcom-M will reduce feed rate and riser temperature. But at steady state, feed rate may return to its setpoint while fractionator overhead pressure may be at a higher value and riser temperature, at a lower value.

Daily averages show the results of reactor/regenerator controls. Fig. 5 presents all manipulated variables except air rate, and Fig. 6 shows the key constraints. Data start the last week in December 1989--6 weeks before the start of commissioning-and continue until the end of December 1990.

Fig. 5 shows the increases in feed rate and riser temperature after commissioning. These increases are caused by computer controls, as well as by changes in feed quality and catalyst.

The 3,000 b/d increase in feed rate attributed to computer controls is achieved by pressure balance, combustion/severity Idcom-M, and unit optimization working together.

Changes in key constraints shown in Fig. 6 are:

Regenerated catalyst slide valve delta-P--from 7.5 psi to between 5 and 6 psi (minimum limit: 5 psi)
Fractionator-overhead throttle valve position--from 45% to between 55 and 60% (maximum limit 60%)
Regenerator pressure--from about 33 psi to 35 psi
Fractionator delta-P--from about 4.5 psi to between 5 and 6 psi (maximum limit: 6 psi).

In addition to the constraints shown, many others are kept closer to their limits by computer controls.

The MV setpoints for feed rate and riser temperatures were determined by the unit optimizer. The economics favored increasing feed rate at the expense of riser temperature.

As a result, most of the time feed was maintained at its maximum limit while riser temperature was maximized as allowed by constraints. This can be seen in Fig. 5 for most of June, July, September, and October.

In April and November, riser temperature was at its minimum limit. At that time, feed rate was maximized against unit constraints.

Note that the calculated optimum solution has not typically been at a "top of the hill" point, where a compromise is made between feed rate and riser temperature and as such, could be implemented by Idcom-M alone without an optimizer.

However, different economic and unit constraints can sometimes result in an optimum operating point that can only be computed by the optimizer.

In this case, the optimizer will result in benefits that exceed control benefits without an optimizer.

REFERENCES

Rhemann, H., Scharz, G., Badgwell, T., Darby, M., and White, D. C., "Online FCC Advanced Control and Optimization," Hydrocarbon Processing, June 1989, pp. 64-71.
Harkins, B., Froisy, B., Amos, D., Frui, D., Nakashima, N., Takahashi, M., "Experience with FCCU Computer Control On-Line Optimization in Japan," 1987 NPRA Computer Conference, Nov. 8-11, 1987.
Froisy, B., Richalet, J., "Industrial Applications of IDCOM," Chemical Process Control CPCIII, Proceedings of the Third International Conference on Chemical Process Control, Asilomar, Calif., Jan. 1217, 1986.
Amos, J. D., Harkins, B. L., Kleinschrodt, F. J., Logue, D. A., "Application of Process Models to On-line optimization," AlChE Spring 1990 National Meeting.
Avidan, A., Owen, H., "Innovative Improvements Highlight FCC's Past and Future," Oil and Gas Journal, Jan. 8, 1990, pp. 33-58.