FCCU ADVANCED CONTROL SYSTEM ACHIEVES 2-MONTH PAYOUT

Per-Olof ErikssonScandinaviska Raffinaderi AB Lysekil, Sweden
Aric TomlinsIcotron Houston
Sachindra KumarDash Icotron Phoenix

Scandinaviska Raffineri AB (Scanraff) fully commissioned a computer-based control system on the fluid catalytic cracking unit (FCCU) at its Lysekil, Sweden, refinery in March 1991.

The project incorporated a number of advanced control applications covering the entire unit.

The application of advanced control resulted in a unit throughput increase, conversion control, optimum product yield distribution from fractionation, excellent product quality control, and reduced operator interaction.

PROJECT BACKGROUND

In May 1988, Scanraff contracted Icotron to conduct a feasibility study for the implementation of advanced control on its FCCU (Fig. 1). The study-completed in September 1988-estimated significant potential benefits, proposed several control functions to attain these benefits, and identified the necessary instrumentation to implement the control functions. With the exception of one item, Scanraff installed all of the recommended instrumentation, and it contracted Icotron in January 1990 to implement the control functions.

The single exception was an instrument on the debutanizer column, which could not be installed because it required a plant shutdown. Scanraff decided to postpone the proposed debutanizer column advanced controls until the instrument could be installed.

Commissioning of the remaining controls began in January 1990 and was completed in March 1991.

CONTROL PLATFORMS

Scanraff uses a Honeywell TDC 2000 platform for the FCCU's basic regulatory control system. The advanced control applications, consisting of control functions and related calculations, were performed on a Data General computer.

Additionally, the applications made use of a data base package and a computer/TDC2000 interface package that were resident on the computer: Texaco Data Acquisition & Control System (TDACS) and Texaco Advanced Control System (TACS). Dynamic Matrix Control (DMC) was used as the multivariable predictive control package.1

IMPLEMENTATION

For convenience, the unit was divided into three sections: the reactor/regenerator, the main fractionator, and the stripper/absorber section. The project was implemented on a section-by-section basis, resulting in three phases.

In the first phase, the stripper/absorber section was completed. The main fractionator implementation followed, and, in the final phase, the reactor/regenerator controls were commissioned.

Although the FCCU is an integrated unit, the targets and constraints for each section were essentially only functions of the independent variables of that section. Therefore, the control applications for the three sections were independent. This simplified the design and implementation of the control functions.

REACTOR/REGENERATOR

The Scanraff FCCU was originally designed to process 132 cu m/hr (20,000 b/d) of hydrotreated vacuum gas oil (VGO). During advanced control implementation, the unit was operating at about 35% higher throughput. The reactor/regenerator section uses an M.W. Kellogg Co. orthoflow converter, which accomplishes all cracking in an external riser. The regenerator operates in a total combustion mode. Fig. 2 shows the process flow for this section.

The plant runs a steady-state optimization program in an off-line mode to determine optimal FCCU operating targets. These targets include conversion, riser outlet temperature, unit charge rate, and fractionation product draw rates.

The unit's operating history showed that the air blower capacity was the primary constraint when maximizing unit throughput. It was believed that the wet gas compressor capacity could also become a constraint.

CONTROL PROBLEMS

The existing basic control system controls the riser outlet temperature by adjusting the position of the regenerated catalyst slide valve (RCSV). The actuator system between the temperature controller and the RCSV is such that the slide valve can only be adjusted in discrete increments.

These increments of slide valve movement are generally not predictable, either in magnitude or when they will occur. On the average, slide valve travel is about 2%. The output from the temperature controller must integrate over a period of time before the slide valve itself moves.

This actuator system makes control of the riser temperature difficult, and the results are generally less than satisfactory. Specifically, the riser outlet temperature always oscillates around its setpoint.

It should also be noted that, as a result of the actuator system, a change in output from the temperature controller generally does not result in any immediate movement of the RCSV.

OPTIMIZATION

The primary objectives for the control system design were to:

Maintain conversion at a plant-specified target.
Maximize unit throughput.
Maintain unit operation within its air blower and wet gas compressor limitations.

Conversion, defined as the ratio of gas oil disappearance to the unit fresh feed rate, is a relatively nonlinear function of the FCCU's operating parameters (or the manipulated variables). Because the DMC uses linear control theory, a different variable--severity--was chosen as a controlled variable.

Severity, defined as conversion/(100-conversion), is believed to be a more linear function of the unit's operating parameters. An Icotron-developed model that is a function of the catalyst type, feed quality, and various operating parameters for the unit, was used for severity prediction.

CONTROLLER DESIGN

The multivariable predictive controller designed for this section included the following manipulated variables and controlled variables:

Manipulated variables (MVs)

RCSV position
Fresh feed rate
Recycle feed rate
Regenerator air flow rate
Disengager/regenerator differential pressure
Wet gas suction pressure.

Controlled variables (CVs)

Riser temperature
Regenerator bed temperature
Regenerator cyclone temperature
Regenerator flue gas 02
Reactor severity
RCSV differential pressure
Spent catalyst slide valve (SCSV) differential pressure
Regenerator air flow controller output
Wet gas compressor suction pressure controller output.

This controller was designed to operate with all controlled variables except reactor severity maintained within ranges defined by a lower and an upper limit. Reactor severity was controlled to an operator-specified target. This strategy ensures that conversion is controlled to a target.

Note that the above multivariable controller manipulates the RCSV position, and the riser temperature is used as a controlled variable. During unit step testing and data analysis, this approach was tested and chosen because the RCSV position could be stepped directly and a first-order-type response observed for the riser outlet temperature.

Using a multivariable predictive approach for controlling the riser temperature instead of the previous proportional integral derivative (PID) algorithm, improves the resulting performance of the temperature control. This is because the effects of the variations in the other MVs (such as the feed rate, differential pressure, etc.) are included in a feed-forward fashion.

The alternative design would be to use the setpoint of the existing PID temperature controller as a manipulated variable. Because of the poor performance of the riser temperature controller, this would have been a bad choice for a manipulated variable.

RESULTS

Fig. 3 compares the daily average data for the predicted severity with the measured severity after the above controller was commissioned. It shows that the model prediction is quite good.

It also shows that this function allowed successful control of the plant for a severity range of 1.3 (conversion = 56.5%) to 1.66 (conversion 62.4%).

Fig. 4 compares the unit throughput before and after advanced control implementation. Prior to implementation, the unit throughput was maximized manually by the operator. The figure shows that unit throughput increased approximately 10%, which can be attributed to the use of advanced control.

For both time periods, the ambient temperatures were approximately equal, and the conversion targets for the unit were the same. Because the unit capacity is typically limited by the air blower capacity, increased throughput usually can only be achieved by some improvement in air blower operation.

Fig. 5 compares the excess oxygen in the flue gas before and after advanced control implementation. It shows that after implementation, the average oxygen concentration is lower than it had been.

After advanced control implementation, the unit was run to an oxygen concentration lower limit of about 1%. The ability to operate at a lower oxygen concentration is partially responsible for the resulting ability to increase unit throughput.

Icotron estimates that, as plant personnel gain more confidence with the advanced control functions, they could reduce the lower limit of the oxygen concentration from 1% to possibly 0.5%. This should help further increase unit throughput.

MAIN FRACTIONATOR

The functions of this area are to separate the hot reactor effluent vapor into overhead vapor and liquid (light naphtha) streams, heavy naphtha, light cycle oil (LCO), and decant oil (DCO).

The main fractionator has four heat-removal systems: overhead condenser, heavy naphtha pumparound, heavy cycle oil (HCO) pumparound, and DCO pumparound. Heavy naphtha pumparound is used as the heating medium in the stripper reboiler; HCO pumparound is used as the heating medium in the catalytic polymerization unit and to produce steam; DCO is used to preheat the FCCU fresh feed and to produce steam.

Light and heavy naphtha are mixed at the battery limits. Therefore, an operating specification for their separation does not exist.

Fig. 6 shows the process flow for the main fractionator.

During the majority of the year, naphtha has a higher product value than LCO. At that time, heavy naphtha is drawn from the tower as a product and maximized with respect to a maximum specification in its ASTM 95% distillation point.

In winter months, LCO can have a higher product value than naphtha. When this is the case, no heavy naphtha product is drawn from the tower. It is then desirable to maximize LCO at the expense of the overhead (light) naphtha. A minimum flash point specification then limits the maximization of LCO. Generally speaking, however, the value difference between naphtha and LCO is small, especially compared to that between LCO and DCO.

The value of LCO is much higher than that of DCO. Therefore, LCO is generally maximized with respect to DCO.

Limitations determine how much LCO can be drawn from the tower. These limitations include maximum LCO product pump capacity, minimum decant product flow rate, and a maximum temperature in the baffle section of the tower.

CONTROL PROBLEMS

The bottom section of the fractionator (from the LCO draw on down) was at times relatively unstable to disturbances. This instability was characterized by pronounced variations in column temperatures and LCO stripper level.

It was noticed that this instability was significantly reduced when the fractionator was operated such that the LCO draw was removing all available internal column liquid. The operators had probably observed this and were typically running the tower in this condition.

The condition described above, commonly known as drawing a tray dry, implies that all of the tower internal liquid to the LCO tray is drawn out to the LCO stripper. There is no internal reflux below the LCO draw.

This mode of operation tends to decouple the section of the tower below the LCO draw from the section above, as far as liquid flow is concerned. Any effect that changes the liquid flow in the top section of the tower, such as a change in the overhead reflux flow, has little effect on the tower below the LCO draw tray, because all liquid is drawn out as LCO product. Observing and comparing the tower operation with and without the LCO draw tray being drawn dry led the project team to conclude that the instability might originate in the LCO stripper.

The combined DCO product and pumparound flows supply heat to the LCO stripper reboiler. Due to the absence of any automatic flow-control device in the stripper reboiler flow (or bypass) line, the heat input to the stripper cannot be controlled.

Any cooling effect that increases DCO product flow also increases heat to the LCO stripper. This in turn increases heat to the top section of the tower. Icotron believes that such a recycle effect causes the instability, and its effects are only present when there is internal reflux below the LCO draw tray.

OPTIMIZATION

The primary objectives for the control system design were to:

Control product qualities while maximizing the most valuable product streams.
Maintain the baffle temperature below its maximum limit.
Maintain light cycle oil production below its maximum pump capacity.
Maintain decant oil product flow above its minimum limit.
Optimize the tower-duty distribution such that heat recovery to the process streams is maximized.

CONTROLLER DESIGN

The control functions designed for this section included a multivariable predictive controller (DMC). This function manipulates a set of independent variables to maintain the controlled variables within constraints and/or at targets. A summary of the key independent variables and controlled variables is as follows:

Manipulated variables (MVs)

Reflux flow
Heavy naphtha draw
LCO draw valve position
HCO pumparound flow
HCO pumparound temperature drop
Decant oil pumparound flow.

Controlled variables (CVs)

Light naphtha equilibrium flash vaporization 100% point
LCO equilibrium flash vaporization 0% point
Baffle temperature
Decant oil product flow
LCO product flow

Additionally, the unit feed rate was used as a disturbance variable for the control function.

The linear programming algorithm of the DMC is used to maximize the production of LCO and to optimize heat recovery from the pumparounds to various process streams. This is done by applying proper cost coefficients for the MVs within the controller. This causes the resultant operation to be on the most profitable constraints.

RESULTS

The largest benefit of the control design was realized through the stabilization of the bottom section of the tower. The controller could maximize the more valuable lighter products (LCO and naphtha) by "properly" distributing the heat load between the DCO pumparound, HCO pumparound, and overhead reflux.

This produced an increase in the combined naphtha and LCO product yields of approximately 3% of fresh feed. This increased yield was at the expense of DCO. As a consequence, DCO production was reduced by a similar amount.

Fig. 7 compares the DCO product flow rate before and after advanced control implementation. For both periods, all other unit operation targets were identical.

The upgrading of DCO to LCO and naphtha was largely limited by the maximum temperature-limit constraint of the baffle temperature. This is shown in Fig. 8, which also compares the baffle temperatures before and after the advanced control implementation.

It is seen that the controller pushed the tower to operate close to its maximum baffle temperature limit. It should also be noted that upgrading the DCO did not cause any of the products to go off-spec.

The controller design and tuning was fairly robust. When model identification was finished and the controller was first implemented, heavy naphtha was not being drawn from the tower as a product. Under such operating conditions, the internal liquid flow between the heavy naphtha draw and the LCO draw is almost 100% higher than when heavy naphtha is being drawn.

However, the same controller operated quite satisfactorily at a later date when heavy naphtha was drawn from the tower. In other words, no controller modifications were necessary to change from one operating mode to the other.

STRIPPER/ABSORBER

The function of this section is to separate the main fractionator overhead product into a depropanized fuel gas product and a liquid feed to the debutanizer, of low-ethane content.

The fractionator overhead product enters the stripper/absorber section as both a liquid stream and a vapor stream (via the wet gas compressor). The separation in this section is performed by a system consisting of a reboiled stripper column, two gas absorbers that are connected in series, and a high-pressure (HP) separator drum.

A simplified process flow diagram for this section is shown in Fig. 9.

CONTROL PROBLEMS

The stripper/absorber section is subject to control difficulties that are inherent to its basic process and control configuration. These difficulties arise because of the presence of a number of recycle loops, or paths by which material can recirculate and accumulate in the system.

At Scanraff, this problem often manifested in the form of a buildup of liquid in the secondary absorber. This was specifically true because the hydraulic capacity of the secondary absorber's rich oil flow back to the main fractionator was limiting.

As illustrated by Fig. 9, there are a number of possible recycle loops in the stripper/absorber section. One such loop consists of the quantity of gas that is absorbed in the secondary absorber. This absorbed gas then passes through the main fractionator, the wet gas compressor, the HP separator drum, the primary absorber, and finally returns to the secondary absorber. Another loop consists of the quantity of gas that is absorbed in the primary absorber. This gas, if the HP separator temperature is sufficiently high, can return to the primary absorber directly through the vapor stream from the separator.

A third loop consists of the gas that is stripped out by the stripper column. This gas can be recondensed in the HP separator, only to be returned to the stripper as feed.

CONTROL INTERESTS

The primary objectives for the control system were to:

Maintain a specified C3 + content in the fuel gas.
Maintain a specified C2 content in the debutanizer feed.
Stabilize the stripper/absorber section operations.

CONTROLLER DESIGN

The presence of the recycle loops and the consequent control problems might lead one to believe that an overall, supervisory-type control function is needed for effective control of this section. However, the response times of the CVs of interest, are quite large, as a result of a change in the available handles (or MVs). This makes it difficult to identify process models.

Moreover, even if the identification were successful, one may not be able to run the controller fast enough for it to be effective because of the limitation of the multivariable package and computer resources. This would result in poor control performance. Therefore, an overall supervisory controller was not designers Instead, six individual control loops, which could minimize the interaction between loops, were designed. These six included the debutanizer overhead product (LPG) quality control (i.e., C2 content), and the fuel gas C3 control.

LPG QUALITY CONTROL

This strategy adjusts the hot heavy-naphtha flow to the stripper reboiler to control the C2 content of the debutanizer overhead product. Feed to stripper is used as a feed-forward variable.

Because the analyzer for C2 content in LPG is located in a downstream column, resulting in a significant dead time for this measurement, a predictive control algorithm (DMC) was used.

FUEL GAS C3 CONTROL

A predictive control algorithm is used to control the combined C3/C4 content in the fuel gas by adjusting the lean oil flow to the secondary absorber. Note that, as was the case in the LPG quality control, the analyzer for the fuel gas composition measurement resided in a downstream column.

RESULTS

Figs. 10 and 11 compare the C2 content in LPG, and the combined C3/C4 content in fuel gas prior to and after advanced control implementation. It shows significant improvement in the operating performance of the unit.

In addition, after the advanced control loops were commissioned, the increase in flows in the recycle loops and the subsequent buildup of the bottoms liquid level in the secondary absorber have not vet occurred. The stripper/absorber section has been operating very smoothly.

Fig. 12 compares the stripper overhead temperature before and after advanced control implementation. It shows that, at present, the stripper is operating with a much lower overhead temperature. This is a clear indication that the stripper is not removing as much heavy material as before, and therefore that the potential for recycling material has been reduced.

In short, the designed control loops perform quite satisfactorily, and an overall supervisory control function is not necessary to meet the performance requirement for this section.

ECONOMICS

The benefit realized from the throughput increase and the light cycle oil yield increase is on the order of $4 million/year. This exceeds our original estimate significantly. The project payout time was less than 2 months.

Prior to the commissioning of each section, the operators were trained in the use of computer controls by the project team.

The described phase-wise commissioning with the "simplest control first" approach, was helpful in the training.

Acceptance of the controls by the operators has been excellent.

The uptime for all the control loops has been greater than 99%.