Graham Bird
David Limb
ChemShare Process Systems Ltd.
Cheshire, U.K.
Alok Pandit
Hinditron Computers Pvt.
Bombay, India
Case studies at a major Middle East refinery, atmospheric and vacuum distillation units at Hindustan Petroleum Corp. Ltd.'s Bombay refinery, and simulation of ammonia production at Erdoel Chemie at Cologne, Germany, have demonstrated the function, capabilities, and advantages of using flowsheet simulation to solve actual plant operating problems.
Flowsheet simulation is becoming increasingly recognized as a valuable tool by designers and operating companies in the hydrocarbon processing industry.
Using process simulation, engineers can gain valuable insight into real plant operations, ensure designs are sufficiently flexible, evaluate alternative operating scenarios, and optimize the performance of process units based on predictions from rigorous models.
Flowsheet simulation, using computer aided engineering (CAE), is becoming commonplace in most operating plants today. Effective application of off-line simulation models can significantly increase plant performance and profitability in both the process and utility sections of a plant through increased production (quality or quantity) and improvements in plant energy utilization and related major equipment maintenance.
SIMULATION ADVANTAGES
The advantage of performing flowsheet calculations on a computer are not limited to savings in manpower and design costs. The real advantage is that the results are much more consistent because the physical property estimating procedures and data are consistent throughout the flowsheet.
Variations in parameters are more likely to give meaningful trends because the same design equations and physical property data can be used for both. Other benefits accruing from flowsheet simulation are:
- Designs of others can be checked to ensure that the plant operates safely within defined limits.
- Value-added products can be efficiently produced.
- Equipment for specified operating conditions can be precisely defined, allowing a narrower design margin and reduced capital cost.
- Alternative processing schemes can be evaluated before making a decision on the final design best suited to defined objectives.
- Engineers and operators can be trained in greater depth of knowledge of how the plant operates and reacts to variable operating conditions.
MIDDLE EAST REFINERY
The use of a flowsheet simulation model at a Middle East refinery illustrates the general approach to flowsheet simulation and some of the detailed work that is involved in modeling a plant's process units. Crude characterization, atmospheric tower (and vacuum tower if applicable) modeling, and preheat exchanger train modeling are the important and basic steps in constructing a flowsheet simulation system.
Separate input files are built and tested for each section (tower and heat exchangers). These can then be merged into a composite flowsheet if desired.
Because of interaction of the tower, its side strippers, and pumparounds, the simulator must be capable of modeling all of these units simultaneously. There is also considerable interaction between the tower and the preheat exchanger train.
This means that the simulation program must give engineers the capability of building a composite model including both towers and rigorous heat exchanger performance calculations in a simple flowsheet. All of these procedures were incorporated into ChemShare's Design II program used in the simulation of the Middle East refinery.
The program runs and converges typical atmospheric tower models on an 80386 microprocessor-based personal computer in a matter of a few minutes. A complex, composite model, including tower and heat exchangers with rigorous rating calculations, is accomplished in 30 60 min.
CRUDE CHARACTERIZATION
The first step in creating a realistic mathematical model of a refinery is to characterize the crude oil feedstock as precisely as possible. Characterization involves translating the feed definition in terms of a batch distillation curve (ASTM or TBP) into relative amounts of pseudo components or petroleum fractions.
For successful characterization, the distillation curve should be as recent and complete as possible. It is frequently necessary to manually extrapolate the curve to 100%.
It is also important to divide the whole crude into a sufficient number of pseudo components or cuts and to ensure that each product contains significant quantities of at least seven pseudo components. These requirements are achieved by the CUT command of the program.
In the program, the term cut is the volume average boiling point of the petroleum fraction. Typically, 30-50 cuts are needed.
A second aspect of characterization is assigning properties to the cuts. The most important of these properties, from a basic characterization standpoint, is the gravity of the cut.
A curve of gravity vs. percent distilled is the method used to provide information to enable the program to assign correct gravities to each of the cuts.
ATMOSPHERIC DISTILLATION
Once the crude feed has been characterized, it is possible to build a model of the atmospheric distillation unit. The model is used to estimate the quantities and characteristics of the distillate and residue products.
This must be done before modeling the crude preheat exchanger train because the atmospheric products and pumparound streams are the sources of heat. Such an approach may be used because the conditions of the crude leaving the furnace are generally known with reasonable accuracy.
In order to model any distillation tower, it is necessary to convert the physical parameters of the tower into those used internally by the program when it calculates the heat and material balance for the tower. The principal conversion is from actual trays to theoretical trays.
This requires an estimate of the tray efficiency. A precise value is not needed, however, because an integral number of theoretical stages is used. Drickamer and Bradford correlated 54 refinery tower tray efficiencies with viscosity.
Preliminary simulations showed that throughout the bulk of the atmospheric tower, and three side strippers, the viscosity lies between 0.2 and 0.3 cp. The Drickamer and Bradford chart suggests an efficiency of about 50 60% for this viscosity range. In the bottom stripping zone, the efficiency may be 10% lower.
Having built and converged the atmospheric tower model, it is then normally necessary to adjust certain parameters to obtain a better match between predicted column temperature profiles and product characteristics (for instance, ASTM 5% and 95% points) and actual plant data. Parameters that may be adjusted include the numbers of trays in each column section, the pumparound flows (for which plant measurements are probably less accurate), and product flow rates.
Reliable test-run data based upon calibrated instruments are needed for maximum success.
To develop the heat exchanger network model without having to run an atmospheric tower model for each exchanger in the model, the product data from a run are usually saved in a stream file.
PUMPAROUNDS
Pumparounds provide a major part of the crude preheat. A special technique for de-coupling the pumparound exchangers from the tower model is used to avoid affecting the tower heat and material balance.
A small, nominal liquid side draw, about 10 kg/hr, is taken from the pumparound draw tray. For initial modeling, this may be saved in the stream file with other stream data.
For modeling the pumparound and crude exchangers rigorously in the network, the nominal side draw is scaled up to the actual pumparound flowrate using a stream manipulator module in the program.
PREHEAT EXCHANGERS
When a reasonable match between the predictions of the atmospheric tower model and measured plant data is obtained, the preheat train exchangers may be modeled. The product streams are read from the previously saved stream file.
First, a simple model of the heat exchanger network is built. For each exchanger, the shell-side outlet temperature is specified. The product streams are generally on the shell side, and crude is on the tube side of the exchangers.
The model may involve some thermal recycles, but it is usually quick to converge because there are no complex rating calculations. There are only heat balances.
From the results, crude preheat temperatures achieved at each stage are obtained. Any major inconsistencies are thus highlighted, such as temperature crosses in the heat exchangers.
The data may then be checked for input errors or suspect plant measurements. Pumparound flows or duties may be adjusted slightly to give a better match to the reported crude preheat.
Obviously, any adjustments to data derived from the tower run require reconvergence of the tower, using the revised pumparound flows or duties.
When a satisfactory heat balance has been obtained for each of the preheat exchangers, and the plant temperatures have been matched with as little adjustment as possible, the detailed geometry of each heat exchanger may be determined for further rating calculations.
First, the outlet temperature or duty specification is retained. The rating calculation is used to determine the surface area required to achieve the duty, the area available from the geometry, and the pressure drop through the exchanger.
Then fouling factors are derived from the first step and inserted in the model. The "temperature-out" step is removed.
These steps are done for each exchanger in the network. The results are then combined for the composite heat exchanger network model.
RECYCLE SOLUTION METHOD
Multiple heat exchanger shells in series comprise a thermal recycle because the temperatures for the two (or more) intermediate streams are not known. To solve the recycle problem, one of the two streams is estimated and then determined by trial and error.
Recycle streams must be initialized in terms of temperature, pressure, and flowrate of each component. Refinery flowsheets typically use about 50 components.
To simplify recycle estimates, the advanced Fortran command, COPSTR, is used to transfer data (primarily component flows) from a known stream to the recycle stream.
The normal calculation sequence begins with calculating the atmospheric tower, unless the product/pumparound streams are available from the stream file. Then the crude preheat train exchangers are calculated, proceeding in the direction of the crude flow.
From this calculation, it is readily apparent which unknown streams require an initial recycle estimate. The furnace is the last item to be calculated.
MODEL USE
The models may be used for answering typical day-to day and longer-term operating questions.
For instance, if the proportion of crude types making up the feeds is about to change, what product property changes might occur if relative product proportions remain the same?
If product properties are likely to be unacceptable, how might the product slate be changed, and how might the operating conditions be changed to bring product properties back to within acceptable limits?
Other questions include: Is there any apparent increase in fouling of the exchangers prior to the desalter? And can additional crude preheat be obtained by varying the flow split between parallel shells?
To answer the question on feed change effects on product properties, the complete model can be rerun with new feed-mix data. If the total feed rate is also changed, then product flows should be scaled up to keep them in proportion.
Retuning the model to improve particular product properties requires a certain amount of expertise, just as it does in an actual plant. This expertise will be rapidly accumulated as plant engineers use the model and systematically record the results obtained from the models. Some guidelines, based on fundamental tower characteristics are helpful for accumulating that expertise.
The overall column temperature profile is strongly dependent on the overhead/bottom split. Increasing the overhead by a few percent at the expense of residue will raise temperatures throughout the tower and vice versa.
Increasing any product at the expense of residue will raise the temperature at all trays up to the draw tray. Above that, there is a lesser effect. Product ASTM temperatures will also rise.
Pumparounds provide reflux for the column sections below. The top pumparound is normally determined by the overall heat balance and can only be varied indirectly.
For example, if the bottom product flow is increased and no other changes are made, the overheads will reduce and the top pumparound duty will increase. This should improve the sharpness of fractionation.
If one of the lower pumparound duties is increased, however, the overall fractionation efficiency will decrease. This is the price paid for recovering heat at a higher temperature level.
Below this pumparound, lower product ASTM temperatures would be expected. But above this pumparound, the 5-95% boiling range would tend to be wider.
In order to use the model to monitor fouling, it is necessary to obtain accurate temperature measurements for all streams to and from the exchangers. The feed and product flows and characterizations should also correspond with current operations.
The individual exchanger model may then be run with a "temperature out" specification and fouling at zero. From the required and available areas, the overall fouling factor may be computed as described previously.
After a period of about 1-3 months, the exercise may be repeated to see whether there is any change in the overall fouling factor.
It has to be stressed that this method is only as reliable as the quality of the stream and temperature data used. Fortunately, errors in temperature will tend to be systematic and, despite such errors, it may be possible to at least identify qualitative trends in fouling.
Shell flow splits can be optimized quite easily. This can be done manually by running a series of case studies, or by using the simulation program's built-in optimizer. For this problem, the objective function would be to maximize the feed temperature to the furnace.
BOMBAY REFINERY
Two atmospheric towers and a vacuum crude distillation tower at Hindustan Petroleum Corp. Ltd.'s (HPCL) Bombay refinery were simulated using the Design II program. The results were found to closely match actual operating data of the plant.
The objective of the case study was to raise the crude furnace coil inlet temperature to 240 C. from 220 C. A new configuration with a preflash tower added ahead of the atmospheric tower was simulated with the program to get heat duty-temperature curves for various products that were to be used for heat exchanger network synthesis.
A total of 15 runs were taken to converge all three towers with desired specifications produced by the technical services department of HPCL.
The distillation system was designed to process 50,000 b/d of Bombay High crude oil. A simple process flow diagram in shown in Fig. 1.
Crude oil is heated from ambient to about 220 C. (the furnace inlet coil temperature) in a battery of heat exchangers called the preheat train. The outlet of the furnace is connected to the flash zone of the atmospheric distillation tower.
Multiple side products are drawn off and steam stripped before exchanging heat with the crude in the preheat train. The column temperature profile is maintained by controlling the reflux, the top and bottom pumparounds, and the flash zone temperature.
Bottom product of the tower is heated to 415 C. before being fed to the vacuum tower. The vacuum tower operates at 70 mm Hg absolute pressure.
Again, multiple, unstripped side cuts are drawn off. Reflux is provided by the first side stream and not by the overhead distillate product.
BASE-CASE SIMULATION
The existing configuration was simulated using the flowsheet simulator. Various runs were made to match as closely as possible the simulation results with the actual plant data.
The program provides options to fine tune the model. For instance, the furnace is treated as an integral part of the atmospheric tower, making it convenient to adjust the furnace duty to match the overflash.
Net liquid flow from the draw-off trays were fixed in accordance with plant design specifications. Similarly, different thermodynamic options were tried to precisely characterize the crude. It should be noted that the equations used to evaluate the K-factor and enthalpy are different for atmospheric and vacuum towers.
The results of the base case simulation are presented in Tables 1 and 2. Note that the results match within 2-4% of the actual plant data (the simulated data are within the accuracy of the plant instrumentation). For comparison, the temperature profiles of both the simulation and the actual plant are shown in Fig. 2.
MODIFICATIONS SIMULATED
After a satisfactory base-case simulation, two different modifications were tried, as suggested by HPCL technical services. The main objective was to increase the heat recovery in the preheat train and increase the tower throughput, if possible, without affecting any change in the furnace.
To gain maximum capacity from the multidraw towers, the top product quantity should be as small as possible, pressure should be high, and circulating reflux should be used. As a result, the addition of one more pumparound and the addition of a preflash tower ahead of the atmospheric tower were simulated, using the base case simulation as a foundation.
The addition of a pumparound was rejected outright after only a few simulation runs because the third pumparound could provide only 3 MMBTU of heat duty. It was realized that this level of heat was not sufficient to invest resources in a new heat exchanger and changes in plant piping.
Therefore, adding a preflash tower was investigated in greater depth. The model was changed to accommodate a preflash tower ahead of the atmospheric tower to remove as many light ends as possible.
The simulation showed that a preflash tower operating at a pressure of 55 psia reduces the load on the main tower, resulting in a stable tower that gives more precise end points of various products.
Also, the removal of 600 lb mole/hr of light ends in the preflash tower yields a lower pressure drop through the crude furnace.
In other words, the crude throughput could be increased if the pressure drop were maintained at the level it was without the preflash tower.
Another significant observation is noted from the heat duty-temperature curves of the products. Apparently, the heat exchanger pinch points (points of minimum temperature difference) were shifted to a more favorable position with the preflash tower installed.
It is not possible at this stage to give the exact reduction in heat transfer area because further work is in progress.
But it is expected that the surface area required in the preheat train could be reduced by about 18-20%.
AMMONIA PLANT SIMULATED
Erdoel Chemie, a major producer of bulk petrochemicals, used CAE to model ammonia production at its Cologne complex. The model is used to calculate heat and material balances and for plant data reconciliation.
Processes simulated at the plant include the hydrogenation reactors, the primary and secondary reformers, the high and low-temperature CO shift reactors, methanation, ammonia synthesis, and other associated equipment, including the plant controls.
The ammonia synthesis model comprises more than 70 modules covering the complete range of equipment. The plant is precisely simulated and incorporates the refrigeration system, purge gas vent system, ammonia reactor flow splitter system, and the quench system. Simulation data show good agreement with actual plant operating results.
The Design II simulation program has been used since August 1987 for the calculation of heat and mass balances for individual components and for simulations of other units in the processing complex.
The ammonia production facilities are not implemented in a single, overall model, but into two sub-models. One sub model simulates ammonia synthesis and the refrigeration system. The other simulates energy recovery.
Simulations include a collection of complete plant data, such as mass, composition, pressure, and temperature of as many streams and components as possible. Individual components are also simulated.
If the agreement between individual physical property phase equilibria data does not closely agree with plant data, it is necessary to provide component and mixture properties to the simulation program data base, using a module called ChemTran.
Connection of streams and components to a flowsheet without recycles is simulated. Input data for the recycles are generated from measured data. Simulation also includes definition of controllers and open-close cycles. The simulation also implements in-line Fortran commands for fast setpoint control and special calculations, such as the steam/carbon ratio, for instance.
In-line Fortran is a special feature of the simulation program which Erdoel Chemie found quite useful for passing stream information and auxiliary calculations before or after equipment module calculation. The feature also allows the addition of other users' equipment and models.
Base cases were developed by matching plant and simulated data. Plant and computed data compare well and generally lie within the range of plant instrument measurement errors (Table 3).
Having established confidence in the results from the base case simulation model, Erdoel Chemie carried out a series of case studies to optimize the operation of the plant. The objective was to increase profitability.
Factors affecting profitability were ammonia product rate and energy utilization.
Case studies were carried out for, but not limited to, the replacement of cracker hydrogen gas with natural gas, the effect of varying H2/N2 ratio in the purge gas, and the effect of varying the steam/carbon ratio.
The simulation showed that little was to be gained by either replacing the cracker hydrogen gas with natural gas or by changing the H2/N2 ratio from a typical value of 3.
Simulation showed, however, that by reducing the steam/carbon ratio from 3.47 to 3.37, energy consumption could be significantly improved, with essentially no change in ammonia production. A comparison of actual and computed data is shown in Table 4.
The success in applying the Design II program to the ammonia plant problems prompted Erdoel Chemie to use the program for other plant situations. For instance, a feed forward analyzer has been implemented for steam/carbon ratio control.
The system is also being used for data reconciliation by checking plant instruments for faulty readings. The simulator compares readings provided by the simulator with readings of the actual instruments.
The system is used to determine the feasibility of plant revamps. For instance, the effect of changing ammonia reactors from multibed, axial flow units to a more efficient, radial-flow design was checked.
The system is used for monitoring plant operations. The program provides the plant engineer with complete data on all streams. This information, most of which is not available from plant instruments, is used to monitor the performance of equipment from both safety and maintenance aspects.
For instance, heat exchanger fouling, ammonia reactor catalyst degradation, and primary reformer tube temperatures are monitored.
BIBLIOGRAPHY
- Smith, Buford D., Design of Equilibrium Stage Processes, McGraw Hill, New York, 1963.
- Gary, James H., and Handwerk, Glenn E., Petroleum Refining Technology and Economics, 2nd Edition, Marcel Dekker, New York, 1984.
- Design II User's Guide, ChemShare Corp., January 1988.
- Bird, Graham, "Computer Aided Design for the Hydrocarbon Processing Industry," Norwegian Society of Engineers, Gol, Norway, November 1984.
- Pandit, Alok, "Flowsheet Simulation in Hydrocarbon Processing Industry: Case Studies Using State-of-the-Art Simulator, Design II, NCL," Pune, India, August 1988.
- Goebel, Norbert, "Computer Simulation of an Ammonia Plant," ChemTech, Bombay, India, February 1989.
- ChemTran User's Guide, ChemShare Corp., January 1988.
Copyright 1990 Oil & Gas Journal. All Rights Reserved.
