OPTIMIZATION AT WYOMING GAS PLANT IMPROVES PROFITABILITY

Lynn E. Saha
Amoco Production Co.
Houston

Andrew J. Chontos
Amoco Production Co.
Evanston, Wyo.

David R. Hatch
ChemShare Corp.
Houston

Amoco Production Co. has implemented a computer-aided manufacturing system for on-line optimization at the Painter complex (Wyoming) gas-processing plant.

The system is based on rigorous process modeling techniques using real time data. Early results show significant potential for improving the plant's profitability.

COMBINED DUTIES

The Painter complex NGL recovery and nitrogen rejection unit (NGL/NRU) is located in the Overthrust Belt about 15 miles north of Evanston, Wyo. Completed in September 1987, the plant was designed to process an inlet-gas volume of 240 MMscfd from the Clear Creek, East Painter, Painter, and Ryckman Creek fields with two parallel trains of 120 MMscfd each.

Design product rates are about 2,500 b/d of condensate and 29,000 b/d of NGL. A schematic of the Painter complex is shown in Fig. 1.

For maximum liquid hydrocarbon recovery from many of the reservoirs, the reservoir pressure must be maintained near original conditions. The sale of produced gas and liquids creates a reservoir voidage which is accompanied by a reduction in the reservoir pressure.

The reduction in reservoir pressure causes retrograde condensation in the producing formation. Consideration must be given to fields operating near retrograde conditions to maximize recovery.

Nitrogen is a relatively inexpensive gas that can be injected into the reservoir to replace the voidage created by hydrocarbon fluid production and to maintain reservoir pressure. As nitrogen injection continues, the produced hydrocarbon fluid has increasing nitrogen content.

Nitrogen has no fuel value and reduces the hydrocarbon gas value on a volume basis. The nitrogen must be separated from the hydrocarbon gas before delivery for hydrocarbon-gas sales.

Accordingly, a deep cryogenic process was required to separate nitrogen from hydrocarbon gas. This is an energy-intensive separation process. A significant volume of valuable NGL is extracted in the process. The nitrogen-gas by-product is compressed then re-injected into the reservoir, reducing the purchase of outside-supplied nitrogen.

The complicated separation process used by the Painter NGL/NRU, the increasing nitrogen content of the feed gas, and the variable product prices make the plant a challenge to operate at maximum profitability. The plant receives gas from four fields with varying nitrogen content ranging from 2% to more than 40%.

Field operations can subject the plant to simultaneous, large rate and composition swings-a uniquely difficult situation. Thus, Amoco Production Co. wanted an on-line optimization system for the Painter NGL/NRU and decided to pursue implementation in mid-1987.

The large number of recycle streams and unit interactions in the Painter NGL/NRU make short-cut analysis inaccurate and rigorous calculations difficult. The ChemShare Corp. was developing a new technique which met Amoco's requirements for rigorous calculations in an online environment.

Accordingly, the Painter NGL/NRU became the first commercial installation for ChemShare's ProCAM, an on-line process computer-aided manufacturing system for maximizing daily profitability. (ProCAM is a service mark of the ChemShare Corp., Houston.)

The system incorporates model-based data reconciliation with a rigorous process-modeling system based upon a proprietary equation-based, simultaneous solution technique.

WHY ON-LINE OPTIMIZATION

In general, gas-processing plants produce on-spec products with minimal operating upsets. Typically, this mode of operation is based on some combination of previous operating experience and "after-the-fact" analysis of plant operating economics.

An accurate study may take weeks to determine the optimum operating conditions for a facility as complex as the Painter NGL/NRU. Then the results may be unreliable because the study relied, at best, on calculations or plant simulations using historical average rather than real-time operating data and product prices.

More times than not, the plants are operated either to maximize NGL recoveries or minimize operating costs (compression costs for example) rather than to maximize profitability because product volumes or utility information are readily available for operator assessment.

Amoco decided to install an on-line optimization system at the Painter NGL/NRU because of the incentive continuously to operate at maximum profitability.

The Painter NGL/NRU is a complex facility to operate because of the integration of the NGL recovery and nitrogen rejection operations. The process requires upwards of 40,000 hp of compression depending on the mode of operation, e.g., ethane recovery, ethane rejection, dew point control, etc.

The process can be manipulated to achieve the desired operating mode, but a timely determination of the most economic and efficient operation is impractical.

Such a highly flexible process is inherently complex, making process optimization extremely complex. The advent of the low-cost, high-speed microcomputer and the robustness of an equation-based exact simulator has made real-time optimization of the Painter NGL/NRU a reality.

We now continuously evaluate the mechanical limitations of the plant and the dynamics of the product marketplace. Thus, as the economics of the process change (e.g., market prices, utility costs, contractual obligations, etc.), the optimization system determines the new operating conditions which maximize the plant profit while simultaneously satisfying the mechanical and thermodynamic constraints of the process.

System installation, testing, and debugging began in September 1989.

BENEFITS AT PAINTER

A study demonstrated the potential improvement in operating pretax profit from the on-line optimization system at the Painter NGL/NRU. This study confirmed the increased profit incentive for on-line optimization.

Based on March 1989 operating data and prices, optimization of 40 key plant operating variables showed a potential pretax profit of $3,000/day. These results were for cases in which ethane recovery was constrained to 30% because of contractual obligations.

Results using the higher December 1989 product prices with no ethane-recovery constraint showed in excess of $9,000/day potential pretax profit increase without a plant operating mode change (i.e., without operator intervention) and about $13,500/day if the operating mode was changed for high ethane recovery (i.e., with operator intervention).

March 1989 operating data for the Painter NGL/NRU were retrieved from the plant data historian.

Each of the approximately 550 data points consisted of the average value calculated from 60 samples collected at 1-min intervals. Thirty-three data sets of plant measurements were collected. The data sets consisted of stream compositions, stream flowrates, process temperatures, process pressures, and compressor speeds.

One data set was randomly selected as the basis for the optimization results presented later.

The data set was reconciled against a rigorous simulation model of the process to establish performance characteristics for the process equipment.

Operating parameters associated with the process-control scheme were then optimized against the reconciled model. Representative product prices and utility costs used in this optimization of the process-operating conditions are shown in Table 1.

The optimization showed that an additional $3,000/day of pretax profit could be realized if key process operating variables were modified. About 30% of the increased pretax profit was attributable to the redistribution of feed components to a product stream of greater value.

Thus, condensate recovery was maximized at the expense of NGL recovery.

The major portion of the increase in the operating profit was achieved by reducing the plant electrical consumption. Like most cryogenic process plants, a major portion of the Painter NGL/NRU operating cost is associated with power consumption.

RECOMMENDED ADJUSTMENTS

Optimization reduced the power consumption by decreasing energy requirements for the cryogenic separation process. This resulted in some lost product recovery and revenue which was more than offset by the power savings.

The recommended operating adjustments for optimization were as follows:

Reduce demethanizer pressure by 44 psi.
Reduce demethanizer bottoms temperature by 24 F.
Increase NRU tower pressure by 15 psi.
Reduce propane-refrigerant condensing temperature by 19 F.
Increase recompressor interstage pressure by 65 psi.
Increase demethanizer feed temperature by 20 F.

The optimization was constrained by limitations of the process equipment, contractual obligations associated with product specifications, and the ethane-recovery level.

Improvements in the anticipated plant profitability were achieved without violating any constraints. All recommended changes could have been implemented as setpoint changes from the plant distributed control system (DCS).

An additional data set from December 1989, taken in a similar manner as the March 1989 data, was retrieved from the data historian.

The plant operating configuration and target ethane-recovery level were similar to the March 1989 case. However, the economic parameters affecting the plant profitability had changed. They are shown in Table 1.

The most significant change was the increased NGL product values relative to sales gas. Also, there was a slight improvement in the sales-gas value relative to the nitrogen product value.

The optimization algorithm recommended increasing the ethane recovery to 53% in response to the increased NGL product values. The algorithm was unable to achieve additional recovery due to the limitations imposed by the heat-exchanger configuration associated with the ethane-rejection operating mode.

The following operating parameter changes were recommended:

Decrease demethanizer operating pressure.
Decrease NRU-tower operating pressure.
Decrease condensate recycle.
Decrease temperature from low-temperature chillers.
Increase methane content of nitrogen product to reduce heat-pump compression requirement.

The recommended changes could have been implemented via the plant DCS, resulting in increased pretax profit of $9,000/day. The plant switched to an ethane-recovery mode in early January 1990.

Using the December 1989 data as a starting point, the optimization model was changed to reflect a heat-exchanger configuration conducive to high ethane recovery.

The optimization algorithm increased the ethane-recovery level to 73% and further increased the anticipated pretax profit to $13,500/day compared with the 30% recovery-level base case. The optimization model produced a set of target set points that could have been used for changing from the current plant operating mode to the high ethane-recovery mode.

SYSTEM COMPONENTS

The computer-aided manufacturing system was implemented for the Painter NGL/NRU to perform on-line optimization in near real time. The system serves as a control-room adviser for the plant operators and engineers by analyzing current data to provide information necessary to drive the process to optimum economic conditions.

The program simultaneously analyzes multivariable interactions of the entire NGL/NRU using an equation-based rigorous plant model.

Process optimization at the Painter NGL/NRU involves six principal steps:

Rigorous reconciliation of process measurements
Rigorous mathematical model of the process
Determination of mechanical constraints
Economic profit equation
Multivariable optimization analysis
Implementation via DCS

After determining the plant is at steady state, the mathematical model of the process must match the plant at the current operating point. To accomplish this matching, data reconciliation is performed because normal measurement errors produce inconsistent data which violate material, equilibrium, and thermodynamic relationships.

The objective of data reconciliation is to determine the set of process data which satisfies the material, equilibrium, and thermodynamic relationships for the process.

The rigorous reconciliation procedure determines the process data using a rigorous plant model so that differences between calculated and measured values are reduced, subject to all mechanical and thermodynamic constraints of the process.1

During this procedure, gross instrument errors are detected and eliminated from the measurement set. The gross instrument errors are flagged and reported for operator or maintenance attention.

Also, during reconciliation heat exchangers, compressors, distillation columns, and other equipment are rigorously rated to determine their performance for subsequent use to ensure optimization predictions remain feasible.

More than 550 measured values from the plant's DCS are continuously polled and reconciled.

An exact or rigorous mathematical model of the entire process is critical to evaluate the multivariable interactions in the Painter NGL/NRU. Short-cut models or partial simulations do not have the accuracy and are often based on engineering estimates or empirical correlations that are valid only for a narrow range.

In addition, the traditional sequential modular approach of modeling one piece of equipment at a time lacks the robustness necessary for online optimization. These obstacles are overcome by simulating entire processes with exact models solved simultaneously through an equation based technique.2

The Painter NGL/NRU optimization system rigorously models 170 pieces of equipment including distillation columns. The rigorous modeling provides an evaluation of the mechanical constraints of the process.

The profit equation used to optimize Painter NGL/NRU operations is simply a mathematical representation of the pretax gross profit of the plant. An accurate profit equation is also critical for achieving optimum plant operating conditions.

Examples of factors included in this equation are feed and product prices, utility costs, and contractual obligations.

After the process data have been reconciled, mechanical constraints verified, and the components of the economic profit equation established, the plant optimization takes place. A multivariable interaction analysis seeking the conditions which maximize the economic-profit equation is performed.

To do this, the process variables to be manipulated for optimization are identified with their feasible limits. These limits, and the contractual, marketing, and mechanical constraints, are necessary to restrict the optimization to operationally feasible regions.

The optimization analysis simultaneously manipulates 40 process variables at the Painter NGL/NRU to determine the optimum operating conditions.

Once the optimum operating conditions are determined, they are reported to the plant operator to be implemented via the DCS. Thus, the benefits inherent to distributed process control are enhanced by overall process unit optimization at the Painter NGL/NRU. The possibility of closing the loop in the future exists.

HARDWARE, SOFTWARE

The on-line optimization system at the Painter NGL/NRU consists of two additional computers added to the existing DCS, a supervisor and a server, each with its own monitor. The supervisory computer is networked between the DCS and the server.

Three separate software programs reside on the supervisor. An interface program polls the DCS and brings real-time process information to the supervisory computer.

The interface program can also return data to the DCS such as setpoints, alarm flags, and status messages. The plant-management program accesses data from the DCS interface program and can return calculated values back to the interface program. The plant-management program also has a data historian package which allows nearly real time data-acquisition and storage.

This program also interfaces with the plant operators and both local and remote engineers through a series of process graphics and menu-driven screens on the supervisory monitor.

The executive program, also on the supervisory computer, was developed to provide the interface between the plant-management program on the supervisor and ProCAM on the server.

ProCAM is installed on the server because of its computational speed. Approximately 2,000 rigorous plant simulations are performed daily on this machine with about 40% of the computational time spent making thermodynamic calculations to ensure precise fluid characterization.

OUTPUT

The optimization system is an operating tool for Painter plant personnel. Accordingly, the supervisory monitor is prominently displayed in the plant control room with easy access by operating personnel.

There are approximately 70 output display screens that the operators and remote users can access as required.

OPTIMIZATION RESULTS

Typical optimization results are displayed on the OPT screen (Fig. 2).

The bottom number represents the incremental pretax gross profit to be made by operating the plant at the optimum state. It is the potential gross pretax incremental profit as defined by the plant profit-equation.

The objective is to make this bottom line as close to zero as possible, indicating the plant is operating at the optimum state. This screen also shows the relationship between operating costs and product revenues.

The operator can clearly see that maximizing product revenue does not always maximize plant profit because there are also potential operating cost increases, in this case, compression costs. Generally, the gas-processing plant will be operated to either maximize revenues or minimize costs.

With this tool, the Painter NGL/NRU has an opportunity to operate at the true optimum state on a continuous basis.

SETPOINTS

Recommended optimum setpoints for the key process controllers are displayed on the SP screens (Fig. 3).

There are 40 process variables at Painter NGL/NRU which affect the plant economics sufficiently to be included in this list. The corresponding controller setpoint changes to achieve the process-variable changes for optimized plant operation are listed.

This is an important information screen for the plant operators and engineers. It shows the controller tag numbers, the initial time-averaged measurements, the current controller setpoints, and finally, the recommended optimized controller setpoints.

RECONCILIATION

When the reconciliation and optimization steps are completed, the results are presented on identical process graphic screens (Figs. 4 and 5).

The reconciled data screens are designated "RD" and the optimized data screens are designated "OP."

There are 31 RD screens and 31 OP screens for the Painter NGL/NRU.

The identical sets of graphic displays show the complete process flow and equipment diagrams similar to the graphics which the plant operators and engineers are familiar with on the DCS monitors.

For each of the more than 550 measurements from the DCS, the RD screen shows the 1-hr time-averaged measured value (top number) and the calculated reconciled value (bottom number),

Similarly, the OP screen shows the current measured value (top number) and the calculated optimum value (bottom number).

The RD displays are useful to engineers and instrument technicians for analyzing the differences between the measured and reconciled process values. The reconciled process values should be considered the "real" process measurement because they satisfy all rigorous model requirements.

Instrument measurement and calibration errors become apparent. These numbers are used for detecting instrument malfunction and have a reliable record for discovering instrument calibration errors and malfunctions.

The OP screens, on the other hand, are most useful for the plant operators because at a glance they can see where the plant is operating and where it should be headed to achieve optimum economic operation.

Additionally, the process graphic displays, both RD and OP, show the controller tag number and current setpoints inside the instrument tag box.

ECONOMIC DATA

Representative Painter NGL/NRU economic data are displayed on the "Ecodata" screen (Fig. 6). These numbers are currently manually input as the product prices and utility costs change.

Amoco is investigating plans for automatic electronic update of these data.

OPERATING CONSTRAINTS

Typical operating constraints are displayed on the CONST screen (Fig. 7). Items such as the base temperature and base pressure (for flow measurement), product constraints, and specifications are manually entered. These generally do not change once entered because they are contractual.

Mechanical constraints such as maximum compressor discharge temperature or maximum vessel pressure are addressed in the general input section of the software and can be changed as required.

Ethane recovery is one constraint which does change at the Painter NGL/NRU on a routine basis. The ethane-recovery level can be set by contractual obligations, although economics may dictate a more profitable recovery level.

In this situation, the target ethane recovery is manually input, and the program will optimize the plant at the given recovery level. Alternately, if the ethane recovery is not specified, then the program will determine automatically the optimum recovery level.

CHROMATOGRAPH DATA

On-line chromatograph data are displayed on the COMP screens (Fig. 8). The appropriate streams are continuously sampled, analyzed, and reported.

The compositions are reconciled to satisfy the plant material and energy balances, then used in the online optimization program. Compositions can also be manually input if there is an on-line analyzer failure.

The data reconciliation step is a thorough plant performance test. The on-line optimization system developed for the Painter NGL/NRU performs numerous plant performance tests every day with absolute rigor. In doing so, all equipment is rigorously evaluated, including shell and tube exchangers, plate fin exchangers, compressors, and distillation towers. Accordingly, a multitude of information is available. With the proper commands, information such as heat exchanger coefficients, heat exchanger fouling, and compressor efficiency is available and can be trended with the data historian feature.

UNIQUE SITE

To the best of the authors' knowledge, the Painter NGL/NRU is the only plant using on-line models as complete and accurate as those used for design of the plant to operate at optimum conditions every day.

The on-line optimization system at Painter is being thoroughly monitored for accuracy.

Many of the tangible benefits of the system have been addressed here.

New and innovative ways of using the vast array of computed data are being discovered weekly. Benefits such as predictive maintenance and maintenance by exception are already being put to test.

ProCAM is able to locate plant bottlenecks. With analysis by the engineer, these bottlenecks can be eliminated and profits further increased. A similar system is currently being installed at the Anschutz Ranch East NGL/NRU; installation at other facilities is under evaluation.

REFERENCES

Nair, P.K., and Iordache, C.D., "On-Line Reconciliation of Steady State Process Plants, Applying Rigorous Model Based Reconciliation," AICHE 1990 Spring National Meeting, Orlando, Mar. 18-20.
Canfield, F.V., and Clemmons, J.U., "Maximizing Process Plant Profitability Through On-Line Optimization, Real Time Rigorous Solutions for Optimal Decisions," AICHE 1989 Spring National Meeting, Houston.