Multivariable control benefits Alberta plant

Kevin Collins
Mobil Oil Canada
Olds, Alta.

Steven Treiber, John Walker
Treiber Controls Inc.
Toronto

• Harmatan production [19452 bytes]

Installation of multivariable constraint controllers at the Mobil Oil Canada Harmattan gas plant has improved plant profitability by increasing the yield of the most valuable products, reducing energy consumption, and creating a more consistent operation.

Mobil Oil installed Optimum Predictive Control (OPC), developed by Treiber Controls, Toronto, on Harmattan's LPG fractionation system (Fig. 1 [19452 bytes]).

The Mobil Harmattan facility, built in 1963, is near Olds, Alta., northwest of Calgary. The main process units at the site are: inlet compression, refrigerated lean-oil absorption, NGL fractionation, sales-gas sweetening, and Claus sulfur recovery.

The facility's production is listed in the accompanying box.

Harmattan installed a Bailey INFI-90 distributed control system (DCS) in 1990 which replaced the existing single-loop control analogue controls. Generally, DCS provides users with a vast quantity of real-time information, but it is not obvious how to use this additional information to improve plant profitability.

An advanced controller must be used successfully to control a rich oil fractionator, a fractionation train, or a sulfur and amine unit. Multivariable constraint control optimizes overall gas plant control by handling both interactions of process variables and time delays in process systems while at all times satisfying process and product specification constraints.

OPC is a technology for the design and implementation of multivariable constraint controllers that operates directly within the DCS system and requires no separate computer.

The addition of advanced control is needed to exploit fully the data handling and computational power of modern DCS. Advanced control, used with a DCS, can increase the profitability of an operation.

Profits arise from improving product qualities (operating closer to product specifications), reducing energy costs, maximizing production rates, or reducing plant upsets. Typically, investment in an advanced control application will be repaid within 12 months.

Opportunities

Four major areas of opportunity exist in a gas plant:

• Sulfur and amine unit

• Cryogenic unit or lean-oil absorption system

• Fractionation plant

• Condensate stabilizer, in plants that have significant liquid yields.

In the present market, the major economic opportunity occurs generally in the fractionation plant. Economic benefits arise from maximizing the production of the highest valued liquid product, while maximizing plant throughput.

The cryogenic unit is usually the next most profitable opportunity for operating improvements. Again, the incentive here is to optimize the separation of methane from the more profitable butane, propane, and natural gasoline (condensate), and to maximize throughput.

In plants with significant liquids yields, the objective is to maximize gasoline product by operating at the Rvp limit.

Finally, the objective in the amine unit is to maximize H2S and CO2 slippage to the sales gas and lower circulation rates and steam consumption.

An economic opportunity in the sulfur unit may arise when regulatory requirements would force a plant shutdown or major capital expenditures; advanced controls can help reduce emissions.

Optimum control

OPC is an integrated software package for the modeling, design, and implementation of multivariable controllers. The design of the software makes it easily implemented on any system which can add, multiply, allow vector operations, and has memory or disk capacity into which a control file can be stored (essentially all DCS).

Efficient use of computer memory and computational speed make this technology easy to implement in any DCS, programmable logic controller (PLC), or computer system. This degree of portability allows for a wide range of applications on different types of computers at the same site and future migration of OPC applications as new generations of instrumentation and computer equipment become available.

OPC provides the user with a "friendly" design environment. Model identification and control design are carried out on personal computers regardless of the type of DCS system being used for on-line control.

The OPC controller is designed with a dynamic simulation of the process. All the design software is centered on graphical output, rather than tables of numbers, and allows analysis of a problem and design of a solution in engineering terms rather than mathematical forms.

The software has been run on several systems: Honeywell TDC 3000, Bailey Net/INFI 90, Foxboro I/A, Honeywell PMX, IBM/RPMS, VAX, PDP11, Moore Apacs, and WonderWare with Windows 3.1+ on PC.

The algorithm

The OPC package generates statistical models of the process that correlate the behavior of measured variables (controlled or constraint variables) vs. manipulated variables.

OPC uses these internal models of the process to anticipate future process responses and plans its control action into the future to provide smooth and intelligent control.

The program's inherent optimization capability permits the control of the process to specifications. It does this while maximizing in an amine unit, for example, raw-gas throughput and CO2 slip to sales and minimizing amine circulation and reboiler-energy consumption.

OPC uses explicit step weight, dynamic models to calculate a strategy which minimizes error across all future control intervals out to the process steady-state. From a technical implementation perspective, OPC uniquely allows the user to design the controller in a friendly environment.

Further capabilities include:

OPC runs equally well in both peripheral computers and DCS (and PLC) systems.

The modeling software provides user-friendly feedback to indicate the quality of the models being identified and provides guidance in how to improve the models with the existing data.

The OPC algorithm has only one tuning parameter for each variable in the control problem. This simple structure does not compromise on performance and yet provides a clean maintainable application.

When actuators fail or the plant operator chooses to take manipulated variables away from the controller, the controller can continue to operate the process while knowing that it is not permitted to move the variables now in the control of the plant operator. This results in control applications with high on-line time.

The design features of the OPC algorithm yield a robust multivariable controller. Actual field experience indicates that changes in gain, time constants, and time delay can be tolerated without changes in controller tuning or models.

Properties estimation

On-line product properties estimation is software used to calculate product-stream properties.

The software uses regressed relationships to predict such product properties as composition, Rvp, and density. On-line analyzers or lab measurement are used to update the estimates.

This approach is particularly effective when analyzer cycle times are slow, analyzers are positioned far downstream of the process or are unreliable, or when no analyzers are available.

The main benefits of this approach are an increase in the reliability and availability of measurements for control. This leads to controls that deliver nearly 100% up time.

The regression relationships are derived from plant operating data using the modeling software in OPC previously described. The regressed relationships are tuned visually with a PC-based program that allows comparison of actual and predicted trend plots.

The on-line software has been implemented in DCS and computers: Bailey INFI-90, VAX, PDP11, IBM/ACS, Honeywell PMX, Honeywell TDC3000 AM, and FOXBORO I/A AP.1

Harmattan project

The major objective for implementing this project was inadequate and inconsistent control of the fractionation train.

Changes in raw-NGL feed rate and composition were only compensated for by on-line analyzer feedback and operator intervention. If not reacted to immediately, overcompensation usually occurred.

Under these conditions, it was impossible successfully to maximize product yield. In some cases, severe penalties were received for violating product purity specifications. This led to conservative operating margins.

Following are the specific objectives for installing multivariable control:

• Create consistent operation between operational crews and reduce process upsets.

• Eliminate the propane give-away in the deethanizer overhead to sales gas.

• Increase the ethane content of the propane product.

• Optimize propane and butane product compositions based on relative product prices.

• Increase the butane content of the condensate product.

The project began in April of 1994 with the field installation of flow meters and transmitters. Plant testing began in July with the first controllers being commissioned and on-line in September. Total project cost was $250,000.

High-quality liquids

The Harmattan fractionation plant has historically produced high-quality fractionated liquids with few impurities. The process control strategy was manually to adjust set points based on on-line chromatograph component analysis and lab measurements.

For the most part, tower control was accomplished by bottoms-temperature control, reflux was rarely manipulated, and no column pressure changes were made during ambient temperature fluctuations.

Process upsets in the lean-oil absorption system, upstream of the fractionation plant, caused large fluctuations in the feedtank level. To keep the feed to the fractionation plant constant, the feed tank level to feed flow cascade control was broken, the flow manually set, and the surge capacity of the feed drum utilized.

Operations personnel monitored feed tank level, but level alarms triggered drastic changes in the deethanizer feed rate. Little manipulation of distillation column variables was done until analyzer feedback was received 20-30 min later. In some cases this caused off-spec product and upsets in downstream units.

To prevent product specifications from being exceeded, operations were maintained with conservative safety margins. This was necessary to reduce the impact caused by disturbances in feed flow and composition. The result was little product optimization.

Typical product qualities before and after multivariable control are shown in Table 1.

Control strategy

The Harmattan fractionation plant (Fig. 1) consists of a deethanizer, depropanizer, and debutanizer. Depending on the marketplace, the maximum benefit is usually obtained by maximizing natural gasoline (condensate) production.

This plant receives raw NGL feed from the rich oil fractionation unit (ROF). The feed flow from the ROF unit tends to vary significantly between liquid and vapor phases, and the feed composition can also vary.

The Harmattan fractionation plant is controlled with four separate multivariable controllers, one for the deethanizer feed tank (V-312), and one for each column.

The control objective for the feed tank is to keep feed flow to the deethanizer as smooth as possible without violating level and other operational constraints.

The control objective for the towers is to maximize the production of the highest valued product without violating product specifications and operational constraints.

On-line analyzers on all of the streams provide feedback at 30-min intervals. The on-line properties estimation (OPE) software uses tower temperatures and pressures to estimate compositions, thus eliminating analyzer deadtime.

The purpose of the feed drum surge volume control is to keep the drum level between limits and to prevent flooding of the deethanizer tower.

The controller adjusts the feed flow rate to the deethanizer to maintain the drum level between high and low level constraints and column differential pressure within acceptable limits.

The ultimate effect of this controller is to smooth the feed flow rate to the towers and permit the level to surge in the feed drum to keep the fractionation train feed relatively steady.

The feed drum control matrix is shown in Fig. 2 [33479 bytes].

Each fractionation column controller was designed to control the impurity in the overhead and bottoms products.

For the overhead product, the heavy key component is controlled and minimized. The light key component in the bottoms product is controlled and maximized subject to compositional constraints.

In this case, the bottoms composition is controlled so as to push the concentration of the impurity up against the product specification constraint.

Fig. 3 [48757 bytes] shows the deethanizer control matrix. Controllers for the depropanizer and debutanizer are similar.

The tower controllers use reflux and hot oil to control overhead and bottoms compositions, respectively. In addition, the reflux is used to keep the accumulator level within acceptable limits.

If overhead condenser duty is lost, the controller will give up on the overhead composition and manipulate reflux to satisfy the accumulator level constraint.

The matrix shows as well that the deethanizer controller also respects numerous other constraints. Violation of these constraints can lead to operating problems, and even when the plant is badly upset the controller will try to stabilize the plant by staying within the constraint limits.

The specific objective of the deethanizer controller is to minimize propane giveaway in the towers overhead and to maximize the ethane content of the bottoms stream subject to the depropanizer overhead ethane composition constraint.

The depropanizer controller was designed to maximize production of either propane or butane as dictated by market price. Product optimization occurs by simply changing the setpoints on the overhead and bottoms compositional targets.

When butane is the higher valued product, propane content in the bottoms is maximized subject to the butane product's Rvp specification. When propane maximization is desired, the overhead compositional target is adjusted so that butane is sent overhead into the propane product subject to compositional specifications.

The debutanizer controller's objective is to maximize gasoline (C5+) production subject to the pipeline specifications.

Benefits at Harmattan

Multivariable control has reduced product compositional variability which has allowed operations personnel to maintain setpoints closer to product specifications.

Increased profits have been realized by maximizing the quantity of a lesser-value component in a more valuable product. In addition, energy consumption has been reduced because products are no longer over purified.

The deethanizer controller has decreased propane loss in the overhead from 1 mol % to 0 mol %. The ethane concentration of the propane product has been increased from 1 vol % (liquid; LV %) to 3 LV %.

No energy savings have been seen in this tower because of the higher reflux rate necessary to recover the propane.

The depropanizer column was severely overpurifying both the propane and butane products through excessive tower reflux and reboiler duty.

Since the OPC controller has been on-line, tower reflux has been reduced by 15% and reboiler hot oil duty has dropped by 30% as the propane content in the bottoms was increased.

To date, the propane content of the butane product has been increased from 0 LV % to 4 LV %.

Debutanizer benefits consist of increasing the butane content of the condensate (C5+) product from 4 LV % to 6 LV %. As a result, tower reflux and reboiler duty have been decreased by 17% and 20%, respectively.

The total economic benefit realized from product optimization and energy savings has resulted in a project payback period of less than 1 year.

The application of multivariable constraint control to this process has exploited the full potential of the Bailey INFI/90 DCS.

The controllers described here all operate completely within a Bailey INFI-90 system. The four advanced controllers, the feed drum, and the column controllers were implemented in about 3 months (testing, design, and commissioning).

The initial project investment was recaptured, as stated, in less than 1 year, and half of this cost was for field instrumentation and tying these process measurements into the DCS.

Additional advanced control applications are now being planned for Harmattan's sales gas sweetening plants and lean-oil absorption system.

Reference

1. Lines, B., et al., "Polyethylene Reactor Modeling and Control Design," Hydrocarbon Processing, June 1993, pp. 119-24.

The Authors