You are on page 1of 11

THE LP AS UTILITY SYSTEMS DESIGN TOOL

Thomas Phillips (thomas.phillips@aramcoservices.com) (khogeer@hotmail.com ). Saudi Aramco. Houston, TX. & Ahmed Sirag Khogeer

ABSTRACT
Petroleum refining is a capital-intensive business. The cost of building a new large complex refinery can far exceed $10 billion in which utilities and offsites compromise more than a third of that. Typically, an LP model of a grassroots refinery project would only be used in the initial configuration studies. After that was done, the LP would be quietly retired early in the Front End Engineering Design (FEED) and only occasionally resurrected to answer the inevitable what if questions. Once FEED was well underway, overall refinery material and utility balances would be performed not with an LP, but with spreadsheets. This resulted in only a limited number of balances for select points in the operating envelope. Although this approach was common in the past, increased computing power and enhanced LP software has allowed the LP to keep pace with and even lead the FEED work. By continuously updating the LP model with the latest unit yields, blending properties and utility consumption/production figures as developed by the licensors and the FEED contractor, the LP can now quickly produce unlimited balances for the overall refinery as well as for individual units. This allows a complete assessment of the operating envelope which in turn leads to enhanced knowledge of and better designs for OSBL utility systems, including steam, power, water and hydrogen. Simulation runs were carried out using ASPEN PIMS to solve the optimization problem. The optimization was done using FEED-quality data developed by the process licensors and the FEED contractor.

I
I.1.

INTRODUCTION AND LITERATURE REVIEW


BACKGROUND

Petroleum is one of the most important sources of power today. In addition, it is the only source of many petroleum-based chemicals known as petrochemicals. The refining industry has suffered considerably during the past two decades. The last US grass roots refinery was built during early 1980s I[3]. This is due to increasingly low refining margins and strict environmental regulations. Many small refineries were forced to quit business since upgrading small and old refineries is usually not economically feasible. In 8 years, the US refining industry has decreased by 40 refineries, from 192 to 152 I[1]. In particular, smaller refineries with zero conversion have closed. Crude capacity has grown by 8% and conversion capacity has grown by 5%. Although 5% seems small, it represents a significant addition of 460 KBPD of new conversion capacity. In such tight situations, a new grass roots refinery should be a large throughput refinery, usually greater than 300 KBPD with a high complexity factor, or even a full conversion refinery, having the ability to produce high quality products for different markets, and preferably being integrated with a petrochemical complex.

I.2.

LP IN REFINERY OPTIMIZATION & DESIGN

Dantiz introduced Linear Programming (LP) in 1947 I[2]. In LP, all constraints and objective functions are linear. One of first principal uses of LP was to solve refinery optimization problems. In 1956, Symons used LP to solve a simplified gasoline-blending problem. Since then, LP has found wide acceptance in refining industry. In planning and scheduling, the constraints are linear (supply and demand). The objective function is also linear, and if not it can be linearized with acceptable accuracy for simple supply chain problems I[4]. On the other hand, for refining industry process modeling, the relations describing the conversion and treatment processes are nonlinear by nature. The scope of refinery optimization LP does not cover heat integration, and usually does not cover steam and hydrogen management. An exception is cases in which the optimization of such items is related to a change in operating conditions or parameters such as flow, operating temperature, type of feed, diverting one or more intermediate products from one unit to another, or in a refinery-wide optimization model (such as utilization of hydrogen from the CCR unit instead of producing fresh hydrogen in the PSA unit). The model that was the inspiration for this paper is unique in that it contains extensive modeling for all major utilities as well as hydrogen distribution and management. Heat integration requires hardware management, such as changing the heat exchanger train configuration. The change of the cooling fluid type or waste heat recovery (such as recovering heat from an FCC unit) is usually a design problem requiring mechanical work. However, this is usually expensive and is beyond the scope of regular optimization models which optimize existing hardware for optimum goals, which is what most refiners are seeking. The Linear Program or LP as it is commonly called is the quintessential tool for understanding overall refinery operations and the complex and numerous interactions between units. In the proper hands, an LP exposes the anatomy of a processing facility and greatly enhances the LP operators understanding of how it works and what makes it tick. When skillfully driven, an LP will bring to light the multi-dimensional operating topography complete with peaks, valleys and the edges of the operating envelope. The LP operator is the pilot who can successfully steer a complex operation through this terrain, whether the facility is up and running or in design. Most refinery engineers are familiar with how an LP is typically used; it sorts through a collection of thousands of refinery variables and finds the combination of these variables that provides the highest margin. In order to do this, the LP must perform precise material and utility balances. The ability to do this quickly and accurately throughout the operating landscape makes the LP a wonderful tool for ensuring that interconnecting piping and utility systems are properly sized and configured.

II THIS WORK
II.1. CONTRIBUTION

The contribution of this work to the field of chemical engineering in general and to the fields of petroleum refining and refinery design is that:

It links the utilities system design to the process system LP, thus the engineer can see immediate results for each operating point in the feasible solution envelope.

It discusses how LP capabilities can be used to expand the role of the LP model in the design of a grassroots refinery project. Given the insight and enhanced understanding made possible with a good LP model it makes sense to use the LP in a project oversight role from initial configuration studies all the way through to startup and hand-off to operations.
It utilizes user-friendly industry standard software in order to have this work accepted and used by the refining industry. Many refining engineers do not use academic research results for actual plant operation.

II.2.

MODEL FORMULATION II.2.1. Objective Function

Utilities are introduced to the objective functions as costs. Thus the LP will try do decrease the cost of utilities as much as possible. The price structure affects the refinery in relation to both feed cost and product revenue. For the feed cost, cheaper crudes may require more processing and treatment to produce salable products and may also be more expensive to ship. The operation of refineries varies with the seasons, and the price structure is affected by the market demand. The operating cost includes catalyst, additives, fuel, power, and cooling water. Most design or optimization models lump all these costs in one category called utilities since the operating costs are smaller in comparison to other items in the objective function. However this simplification can result in misrepresentations of refinery operations and economics and make the LP unsuitable for use in the latter stages of FEED. If sufficient utility detail is included in the LP model, it can be a unique and insightful tool for offsite system design. The objective function can be expressed mathematically as follows for a given optimal refinery operating case that is for a fixed feed and products streams: Minimize = Pt (2)

Where is the cost -based objective function, Pt is the total cost of utilities given by the following equation: Pt= (Prod j. Price j ), (3) where j = 1, ..., 7 is the utility for the refinery, such as LPG, gas, HP steam, etc., Prod j is the consumption (or production rate) of utility j , units can be different for each utility such as KWH for electricity, Tons per hour for steam, Price j is the price of a unit quantity of product j, for example, $/Ton steam. II.2.2. Constraints

In addition to the refinery constraints are usually capacity limits, process variable constraints and product specifications.

Table 1: Typical Refinery Constraints: Unit Capacity constraints Steam Constraint Name Lower limit 160 20 50 Upper limit 250 50 100

Process constraints

HP steam MP steam LP steam Water cooling fresh BFW Steam balance Water balance

These constraints can be written as: Steam total = (Steam k X wt f) MP Steam 0 .18 Steam Total ; MP Steam 0.26 Steam Total etc II.2.3. Termination Criterion

The termination criterion is n+1 - n < , (6) Where denotes the objective function value, the subscript indicates the iteration number and represents the absolute difference in the objective function in two successive iterations. II.3. A DESCRIPTION OF THE REFINERY

The model was based on a 400 KBPD full conversion refinery in the Middle East with the ability to export products worldwide. The refinery has a coker, hydrocracker, CCR, distillate hydrotreater and no FCC. II.4. SOFTWARE SELECTION AND FEATURES

Simulation runs were carried out using ASPEN PIMS. It was preferred to continue the runs using PIMS since it is the primary operations optimization tool for many refineries and its use is more convenient for process engineer. PIMS uses several tools to perform the optimization, including a third party optimization engine known as XPRESS from the Dash Optimization software company. In addition, PIMS uses the same basic chemical engineering equations needed for mass and material balance. This software has the capability of using various techniques to solve the nonlinear parts of the model. These include Successive Linear Programming, Recursion and Distributive Recursion for the global model case.

III IMPLEMENTATION III.1. THE LP TEAM ONE RECIPE THAT WORKS Building and maintaining an LP is a tedious job. Usually, this is done by a dedicated individual working alone. We have found however that the process can be more effective when done by teams of two or more people. Experience has shown that an effective two person team consists of an LP architect and an LP operator. The LP architect knows all the tricks and traps of the LP modeling system. He has the knowledge, patience and perseverance to build and debug LP structure. Ideally, he is also quite knowledgeable of refinery processes and possesses a rich data bank of refinery process data. The LP operator ideally has years of refinery engineering, operations and planning experience and understands the LP structure well enough to make minor structural changes. He thoroughly understands refining, product blending, refinery economics and the capabilities and power of the LP but does not necessarily have the patience or in depth LP knowledge of the LP architect. Ideally, the LP operator has extensive experience actually using an LP to plan and optimize refinery operations. Having such a two person team with these diverse skill sets reduces the tedium of LP model construction and helps create an atmosphere conducive to creating a truly inspired product. The operator and the architect together ensure that the model truly and accurately represents the capabilities of the process facility. Together they can accomplish more than could be done by either one alone. It works like this: After making multiple runs, the operator determines that the LP needs more structure and functionality in order to perform a specific task or better represent a specific operation. Following detailed discussions with the operator, the architect adds the rather complex structure required and after a few debugging runs turns the model back to the operator. The operator then extensively tests the model under various constraint scenarios to determine if this new structure works properly and truly represents what his experience tells him will actually happen in the refinery. Extensive communication between the architect and the operator is required to efficiently grow an LP. As in most work environments, this communication is ideally facilitated by a high degree of camaraderie. This camaraderie in turn is fostered in an environment where the LP is viewed as entertainment rather than work, i.e., a fun video game simulator with the goal of truly understanding the essential process and economic nature of the refinery. When this understanding is achieved it becomes a relatively simple matter to optimize the design of a new refinery or the operation of an existing one. III.2. THE OLD WAY OF SIZING OFFSITES Normally a project LP is shelved shortly after the early configuration work is done. During this configuration phase of the project, generic yield and utility consumption data are used in the LP for the process units. For high level configuration decisions, this degree of accuracy is normally all that is required.

As the project moves into FEED and data is received from the licensors, static spreadsheet tools would then be used by the FEED contractor to perform the material and utility balances in each individual unit as well as across the entire refinery. Because of the complex interactions between refinery units and the engineering labor required to generate the balances, these spreadsheet balances would be performed for a few cases, i.e., normal operations and a handful of off-design extremes. Offsites facilities would then be designed against these few cases. III.3. A BETTER WAY OF DOING BUSINESS In order to optimize the design of the offsite systems, the operation of the refinery should be accurately simulated at 20 or 30 points inside and on the edge of the operating envelope. A team representing both operations and project design engineers should determine these points and then the material and utility balances should be done for each identified case. Ordinary spreadsheets are not up to this task. In addition to their rigidity and labor intensiveness, they cannot accurately deal with product blending and the multiple modes of operation possible in many of the ISBL units. Typically, only one mode is modeled for each unit when in reality a continuum of modes is possible. Consequently a more capable tool is required if a more in-depth understanding is desired. That tool is the up-to-date and accurate project LP. After the early configuration work is done, work should continue on the development of the project LP model. As soon as it is available, yield and utility data from the licensors should be incorporated into the LP. Stream blending data and utility summaries should be requested from the licensors for all base and rated cases. This data must then be entered into the LP. In addition, structure must be added as required to properly simulate the utility generation options under consideration. Some ISBL units have multiple modes of operation with distinct utility profiles for each mode. For example, suppose the CCR is designed to operate both with and without benzene precursors and at 100 and 102 RONc. This would result in four distinct modes of operation with four distinct sets of utility balances and stream blending data, all of which should be obtained from the licensor and incorporated into the LP model. Once data for the four modes is available, others modes can be interpolated or extrapolated from the data. For example, a 96 RONc mode could be extrapolated for both precursor modes resulting in a total of 6 CCR modes in the LP model. By comparison, a typical or conventional utility balance spreadsheet used for offsites design would contain only the base CCR design case (precursors to the CCR, 102 RONc). During the multiple LP runs made to characterize the operation of the refinery, it may become clear, for example, that 102 RONc is not the normal mode of operation. Maybe its closer to 100 RONc if this mode maximizes refinery margins while making on-test motor gasoline product. This lower severity mode in turn changes steam production in the hydrogen plant and steam consumption across the gasoline block. Steady state operation is but one point on the curve, and it is often not the case that sets the design conditions for the offsite facilities. What are these extreme design cases? This can be determined by carefully examining and comparing the utility balances produced by the LP for the many operating cases mentioned above. These cases can include partial and total shutdown of various process units during various motor gasoline blending seasons for both expected and guaranteed hydrogen production/consumption across the CCR and hydroprocessing units.

III.4. UTILITY DATA INCLUDED IN THE LP In order to use an LP for utility design, accurate utility data must be entered and structure added as required to properly simulate the utility generation options under consideration. As a minimum the LP must include the following utility data for each ISBL unit: Electrical power consumption Boiler feed water consumption Steam production/consumption for each level of steam used in the refinery. Typically, this would include separate entries for each ISBL unit for high, medium and low pressure steam Condensate production/consumption Cooling water consumption Fuel consumption/production For the offsites, the following should be modeled in the LP: Steam generation system Power generation Steam letdown stations Steam condensing stations, if any Boiler feed water de-aeration Flash steam production Cooling water distribution All miscellaneous offsite steam and power consumers. These can be modeled as fixed rates or tied to appropriate process variables. Examples of these miscellaneous items include: o o o o o o o o o o o o o o o o o Tankage Flare System Waste Water Treatment & Drains Fuel Gas System ISBL De-superheating sprays Flash Steam Small IOU Steam Drivers Misc. OSBL Allowances & Steam Losses Water Treatment & Supply Cooling Water System Sea Water System Air System Nitrogen System Fire Protection Coke Transport & Storage driven by coke production rate Sulfur Pellet Loading driven by sulfur production rate Plant wide lighting, HVAC & Power Losses Allowance

III.5. INTERPRETING THE DATA Once the utility design cases are defined and run (there could be 30 or more) an MS Excel workbook can be prepared that uses LP produced spreadsheets as input. This workbook can then produce simplified, easy to view summaries for the fuel gas system (including composition and composite heating value data), the hydrogen plant feed system (including rate and properties of each feed stream), and all major utility systems (including boiler duty, break over station duty, LP steam condensing duty, fuel gas, cooling water, BFW, cooling water and electrical power used/consumed, imported fuel requirement, condensate flash steam produced, etc). All parameters required to close the major utility balances can be provided by the LP. A careful evaluation of all cases will then reveal the extreme design points for all major utility systems. Once these various extreme design cases are determined, the FEED contractor can then produce the traditional spreadsheet balances for these identified cases in order to verify the LP output and to document the IOU design decisions. This approach provides a high level of confidence that the offsite systems will not be over or under designed and will be capable of handling a wide range of operating scenarios. This is a degree of confidence and design efficiency not possible with the traditional methods. III.6. CASE STUDY: HYDROGEN OPTIMIZATION AND DISTRIBUTION The same material balance and utility balance strengths inherent in an LP that make it an ideal tool for understanding and designing utility systems also make it an ideal tool for optimizing and designing the hydrogen distribution system. This can only be done if the LP is populated with accurate hydrogen production and consumption data for all hydrogen-involved units, at all levels of hydrogen purity throughout the refinery. Typically, this would include separate tracking of high purity PSA hydrogen, CCR net gas and any and all other low purity hydrogen streams. These lower purity streams would include: Hydrocracker cold low pressure separator gas Solution hydrogen in the Hydrocracker, hydrotreaters and CCR off gas Saturate gas plant off gas Hydrogen distribution is a complex issue in a modern refinery. There are many sources and many destinations. When designing the interconnecting piping for hydrogen distribution, it is important to understand the operating circumstances and the flow rates and composition for each branch of the network. When accurately modeled, the LP can help the FEED engineers more completely understand the dynamics of the hydrogen distribution system. For example, the primary destination of CCR net gas may be a diesel hydrotreater. But how should this stream be handled when the diesel hydrotreater is down? Options may include a PSA unit, the Hydrocracker, feed to the hydrogen plant, the fuel system, and finally the flare. Because of its powerful material and utility balance capabilities, the LP can quickly, accurately, and optimally re-distribute hydrogen under any possible operating scenario. This is a powerful design tool which again gives the system designers a high degree of confidence in the final system design. As was done with the utilities systems, the LP spreadsheet output can be quickly copied and pasted into a workbook to produce tabular and graphic depictions of the hydrogen distribution system, an example of which is shown below:

Figure 1: Hydrogen Distribution Diagram


HYDROGEN FLOWS
BOLD, 3 DECIMALS: ITALICS: BOLD, 1 DECIMAL: PURE H2, MMSCFD HRG, MMSCFD PROCESS UNIT FEED RATE, MBPD 5.056 H2 TO SGP NHT MBPD 81.2 scf/b 207.6 TO FUEL 0.000 19.881 23.144 CLPS GAS 19.881 23.144 2.873 HCK MBPD 115.8 scf/b 1,925.0 From CCR LOSS

0.824 H2 TO SGP ISOM MBPD 14.7 2.031 scf/b 138.4

HYR 0.000 0.000 TO FLARE 0.000

HYH 0.000

220.655 0.000 0.000

2.340 2.527

CCR MBPD 81.7 scf/b 1,788.7 HRG: 157.802

HYR 146.110 157.802

46.591 50.319 TO TGT 0.000 TO HKR 2.527

CCR PSA

HYH 41.000

220.655 5.591 Loss to FG Loss to FG 1.757 HYH 223.655

H2 PLANT FEEDS

25.4 0.0 12.1 0.0 0.0

C1C2 C3 C4

MMSCFD MBPD MBPD

NAPHTHA MBPD
CCR Net Gas

7.636 H2 TO SGP DHT MBPD 169.5 scf/b 573.3 LOSS 4.221

TO FUEL 0.000 TO HGU 0.000 13.519 H2inFG Feed

MMSCFD

HYH 0.000

HYR 97.179 104.955

HGU 170.894 PSA

HYS 182.655

HYH: Hi purity H2 HYR: Reformer H2 HYS: SMR H2 HYH 3.000 0.000 NNF HYR 0.000 0.000

TGT

HYH

III.7. THE NEW PARADIGM: MAINTAINING THE LP THROUGHOUT THE PROJECT LIFE CYCLE An accurate LP, in the hands of a skilled and dedicated team, can quickly and accurately address a multitude of design, operations and economic questions that arise during FEED. In addition to addressing the inevitable what if configuration questions that arise throughout the FEED process, the LP will also provide input into the project economic model, and as we saw above, assist with the design of the utilities and interconnecting piping. As such, it is a tool that should be maintained and utilized throughout FEED. When it is all said and done, LP is the only tool capable of accurately and quickly assessing overall project performance. Maintaining the LP throughout FEED then provides project and company management with key information they need to make final investment decisions. Without a good LP in the hands of a skillful team, the picture is far less clear and less accurate. As the project moves into detailed design, the LP should continue to be updated with design data as it is made available by the EPC contractors. During this EPC phase, all functionality required by Operations to start up and operate the facility should also be added. At some point during the EPC phase, Operations should assume ownership of the LP. Smooth startup and operation of the new facility will be greatly enhanced by skillful use of the LP. If possible, the LP architect and LP operator from the FEED phase will continue to support the LP during startup. If this is not possible, they should at least spend as much time as

possible turning over the LP to Operations before they move on to new ventures. In order to enhance this process, the LP should be designed from the beginning with the understanding that it will eventually be handed over to operations. This means providing adequate operating flexibility (not just rigid, one mode unit models) and copious documentation to facilitate understanding of model structure. Normally, this documentation will take the form of comments and notes inside each unit submodel. As is the case with most LP models, survival of the model is dependent in large part on the how effectively the model is handed over from one team to the next.

IV CONCLUSION & RECOMMENDATIONS


a. The LP is much more than just an optimization tool. In addition to performing its legacy task of finding the optimum configuration, it is also an important design tool that can effectively be used throughout FEED for utilities and interconnecting piping design. b. Maintaining the LP throughout FEED will improve overall project integrity and will improve the quality of project information transmitted to management. c. Effectively transferring the LP to Operations during the EPC phase will facilitate a smooth project startup and maximize overall project profitability. d. It is recommended that this work be extended by combining the LP with off-the-shelf scheduling software to optimize product tankage and blender designs. e. It is the opinion of the authors that essentially all refinery and petrochemical project FEED efforts for both ISBL and OSBL systems can be enhanced by effectively utilizing an accurate LP throughout the entire project life cycle.

V ABBREVIATIONS
Abbreviation CCR CONST DHT FCC KBPD LP PIMS ( or ASPEN PIMS) ISBL OSBL FEED Description Continuous Catalyst Regeneration (Type of Catalytic Reforming) Constraint Distillate Hydrotreater unit used to reduce diesel sulfur content using hydrotreating process Fluidized Catalytic Cracking unit Thousand barrels per day Linear Programming (also known as linear planning) Process Industry Modeling System by Aspen-Tech Inside battery limits units that refers usually to process units Outside battery limits that includes storage, interconnections Front end engineering design

VI REFERENCES
[1] DUNCAN, N. REFINERS CAN BOOST PROFITS USING THE CONVERSION INDEX. OIL & GAS JOURNAL. V98, 2000. [2] LI, X. REFINERY-WIDE OPTIMIZATION. PHD DISSERTATION. TEXAS TECHNICAL UNIVERSITY, 2000 [3] SAUDI ARAMCO TECHNICAL COURSES SERIES ON PETROLEUM REFINING PROCESSES MANUALS, 2000. [4] UOP FCC PROCESS TECHNOLOGY COURSE MANULAS. UOP. 1996. [5] ZHANG, J, ZHU, X AND TOWLER, G. A SIMULTANEOUS OPTIMIZATION STRATEGY FOR OVERALL INTEGRATION IN REFINERY PLANNING. INDUSTRIAL AND ENGINEERING CHEMISTRY RESEARCH. V40, 2001.

You might also like