Computers and Chemical Engineering 26 (2002) 1303– 1319 www.elsevier.com/locate/compchemeng
Refinery advanced process control planning system Haitham M.S. Lababidi a,*, Samir Kotob b, Bader Yousuf c b
a Department of Chemical Engineering, Kuwait Uni6ersity, P.O. Box 5969, Safat 13060, Kuwait Department of Systems and Control, Engineering Di6ision, Kuwait Institute for Scientific Research, P.O. Box 24885, Safat 13109, Kuwait c Kuwait National Petroleum Corporation, P.O. Box 10252, Shuaiba 65453, Kuwait
Accepted 9 April 2002
Abstract The process of enhancing a refinery’s performance through advanced control usually requires extensive analysis on the plant to determine factors that are constraining the plant from reaching optimum production conditions. An advanced control planning environment (APCO) was developed using mixed integer-programming methodology. It was supported by a friendly graphical user interface (GUI), to make it easy and accessible to refinery planning engineers. Advanced process control benefits programmed into advanced process control optimizer (APCO) are based on benefits already computed for a major refinery in Kuwait. In here, APCO has been customized to adapt a simplified flowsheet model for a 60 000-barrel per day refinery. A number of case studies are included to demonstrate the capabilities of APCO. The case studies reflect situations encountered by decision-makers in refineries. These include single and multiyear plans, as well as self-funded and unit capacity limiting plans. APCO facilitates developing and evaluating different economical scenarios. Results show that analysis of APC benefits and their interaction among units would uncover APC implementations that are otherwise not considered by simplified payback period analysis. © 2002 Elsevier Science Ltd. All rights reserved. Keywords: Advanced process control; Planning; Optimization; Refining; Mixed integer programming
Nomenclature Ai Ai Bi Bi BFi Cik Gijk Hik ND NI NS Nu NSi rik Sik Uji
set of all units downstream of unit i amount of available input flowing into source unit i, T/D set of all units upstream of unit i amount of flow required out of destination unit i (based on the product demand), T/D total yield benefit in unit i cost of introducing strategy k into unit i, $/T incremental change in flow from unit i to unit j using strategy k, T/D increment in capacity of unit i due to capacity benefit from using strategy k, T/D set of destination (product) units set of intermediate units set of source units set of units for which APC strategies are applicable total number APC strategies applicable for unit i reduction in operating cost of unit i due to strategy k, $/T binary decision variables to indicate the selection of APC strategies upper limit on feed flow rate to unit i from unit j, T/D
* Corresponding author. E-mail address:
[email protected] (H.M.S. Lababidi). 0098-1354/02/$ - see front matter © 2002 Elsevier Science Ltd. All rights reserved. PII: S 0 0 9 8 - 1 3 5 4 ( 0 2 ) 0 0 0 9 6 - 0
H.M.S. Lababidi et al. / Computers and Chemical Engineering 26 (2002) 1303–1319
1304
Vij Xij X( ij Y( ij i
value of stream flowing from unit i to destination unit j, $/T flow rate of the stream flowing from unit i to unit j, T/D base case flow rate of the stream flowing from unit i to unit j, T/D nominal yield of output stream flowing from unit i to unit j set of mutually exclusive strategies for unit i
1. Introduction Advanced process control and real time optimization have become viable means to enhance refineries and process plant profitability, by increasing process unit capacity, and yield of more valuable products. However, like any other process enhancement activity, several alternatives are available to the refiner or process operator that cover issues such as APC technology, and on what unit to apply it. Quality measurement used for feedback control is another aspect that needs to be explored. In here the issue is whether to acquire an analyzer, or to develop an inferential calculation. In addition to the financial impact, such decision will have an impact on the performance of advanced control in term of repeatability and dead time compensation. Process complexities, non-linearity and long time delays had always characterized most processes encountered in the refining and process industries. Advanced and multivariable predictive controls had become the focus of research to deal with such difficulties. The pioneering work of Richalet, Rault, Testud and Papon (1978) which demonstrated the use of IDCOM for an FCCU distillation column, and a steam generator, had ushered in a new wave of application developments and field trials in refining, process and steam generation industries. An alternative multivariable approach to IDCOM using dynamic matrix representation and known as Dynamic Matrix Control (DMC) was developed by Cutler and Ramaker (1979), Cutler (1983). Following these beginnings which could be characterized as pilot applications of advanced and multivariable control, advanced process control started its penetration of industrial plants, driven by real process operation difficulties, and motivated by promises of economic benefits. Field application of advanced control was reported by Agnihotri, Bourgeois, Crosby and Hammann (1987) on an H-oil hydrocracker unit, Grosdidier and Kennedy (1990) on fractionators, van Wijk and Pope (1992), Paraschiv, Marinoiu, Patrascioiu and Cartoaje (1996) on atmospheric crude units, Mastrogiacomo et al. (1997) on a hydrogen unit, and Korchinski, Black, Li, Warrington and Rasmussen (1992) on an FCCU. Methods of estimation of benefits realized by improved controls were presented by Martin, Turpin and Cline (1991). Estimation of energy saving benefits was reported by Bozenhardt and Dybeck (1985). Richard,
Watson, Danzinger, Tuppinger, Schuster and Wilmsen (1995) reported capacity credit and temperature control benefits on an FCCU pre-heater. Fast payback period for advanced control in a refinery was reported by Bullerdiek and Hobbs (1995). In a survey on applications of control technologies in Japanese industry, Takatsu, Itoh and Araki (1998) stated that the benefit of the investment appear in the form of quality improvement, labor saving, production increase, energy saving, and resource saving. Typical benefits of APC technology were listed by Goodhart (1998) as increased throughput, maximization of more valuable product, reduced energy consumption, steadier operation and improved recovery from process upsets, increase in run length and improved safety, and increased process understanding. Various design alternatives can be addressed by relying on benefit and cost analysis to determine where to invest. Among such alternatives, it is important to determine the alternative that will lead to overall optimum outcome. Such determination is best carried out within the framework of an optimization model, where all process parameters and constraints are factored in. The complexity of dealing with optimization models of large refineries, the amount of data involved and the need to generate and analyze different scenarios, necessitate developing an interactive environment for APC planning. Such environment enables the refining engineer to maintain all refinery-related data as well as analyzing and comparing the optimum APC plans. This paper presents an APC planning system, called Advanced Process Control Optimizer (APCO). APCO is an interactive computer based environment which implements a decision model for a typical refinery. The system provides solutions to the following decisions: finding what advanced control strategies to apply, determining which units to apply the strategies to, establishing any interaction in benefits that may arise between units, addressing existing and future operational and market constraints that could have an impact on the refinery’s profitability, and planning such investments over a multi years horizon. The performance of the planning model is demonstrated on a simplified refinery flowsheet consisting of nine units. Data related to APC benefit calculations
H.M.S. Lababidi et al. / Computers and Chemical Engineering 26 (2002) 1303–1319
and implementation costs is derived from an actual refinery study. By applying the planning model, it is possible to generate comprehensive advanced control investment decisions for a plant. Typical APC benefits that were derived from observed refinery benefits are discussed in the next section. Development of the APC planning model is then outlined followed by the design and implementation issues of the planning environment, APCO. A number of case studies are included to verify the performance of the planning model and to highlight the main features and capabilities of APCO.
2. Control benefits and priorities Advanced process control benefits in a unit may include one or more of the following forms: increased throughput. Increased yield of more valuable products. Reduced quality giveaways. Reduced energy consumption. Reduced reprocessing cost. Improved process stability. The above benefits may or may not materialize in every unit. Yield pattern change of each unit is quantified if it is economically or operationally significant. The interaction between unit benefits needs to be determined accurately. Close review of upstream and downstream conditions that can have impacts on the computed benefits needs to be carried out. Typical APC benefits were reported by Hall (1985). Humphrey and Seibert (1992) argued that advanced control projects have better than 25% per annum rate of return on investment. Proposed APC benefits are elaborated in the following paragraphs.
2.1. Increased throughput (or capacity credit benefit) Friedman (1994) computed a 3% throughput increase on crude from an APC system for a crude unit. Rivas and Barsamian (1989) reported very high throughput increase in a FCC unit. Similar results were reported by Eriksson, Tomlins and Dash (1992). It is well known that increases in throughput are the most important type of benefits in term of financial returns to a process from advanced control. Such increases in any unit’s throughput, even if small in magnitude, and if not limited by the capacity of upstream or downstream units, will translate to several million dollars annually. Conversely, APC benefits that do not include increased throughput benefits will remain limited. Strategies that are known to lead to increased throughput are constraint-pushing strategies, where feed is continuously pushed until a certain physical constraint is reached.
1305
2.2. Increased yield Friedman (1994) cited an increase of 1.5% in yield of white products on crude. Davis, James, Catt and Adair (1986) demonstrated significant improvement in product recovery and improved yield by using advanced process control. This is a common type of benefit, where an advanced controller, such as a simple quality control loop, or a MVC could directly affect the performance of a process unit. In this type of APC benefit, the yield of more valuable product is increased, while the yield of less valuable products is reduced. The most used tool to gain the benefit of yield increase is the multivariable predictive controller such as DMC™ and (Robust Multi-Variable Predictive Control Technique) RMPCT™ when applied to a crude unit, or multi-draw fractionator. An example of this benefit in the Kuwait refinery was the case of residue desulfurization unit distillate, where its yield was increased and the yield of bottoms and naphtha were decreased.
2.3. Reduced quality gi6eaway As in any industry, manufactured products should meet several specifications. In refining, it is not uncommon to have conflicting requirements, where meeting of one limiting specification could lead to quality giveaway in another. Examples of limiting properties encountered include viscosity for fuel oil, sulfur content for gas oil and octane for gasoline.
2.4. Reduced energy consumption Typical energy savings in plants from APC are covered in Pelham and Moriarty (1985). Additional savings could be realized by on-line optimization as shown by Wellons, Sapre, Chang and Laird (1994), where more than 3% in fuel consumption was reduced in the power plant of a Texas refinery. This benefit refers mostly to heaters and heat exchangers, steam boilers, and steam reformers (Mastrogiacomo, Bilodeau, Treiber & Walker, 1997). In a crude unit, the application of APC related benefit to energy savings were estimated to be a reduction of 5% in fuel consumed per cubic meter of crude processed (Friedman, 1994). Energy related strategies are similar to those controlling excess oxygen, or temperature and load feed forward. The benefit materializes from operating at higher temperature, or lower excess oxygen.
2.5. Reduced reprocessing cost Combining this benefit with the energy cost benefit constitutes the total operating cost benefit. To distinguish this benefit from energy cost benefit, this benefit refers to savings in utilities such as steam, and hydro-
H.M.S. Lababidi et al. / Computers and Chemical Engineering 26 (2002) 1303–1319
1306
gen consumption or production. Strategies that lead to this type of benefit are stripping steam ratio control on crude tower strippers. Reaction severity control loops for reactors have a similar impact. Steam to hydrocarbon ratio control will lead to energy consumption in hydrogen production. From practice, such benefits may not be economically significant, however, they may be operationally critical. For instance, if hydrogen supply is limited, any severity control benefit could manifest itself in less product giveaway, more optimal yield, but most importantly, it may relax some constraint on other refinery units leading to net capacity credit.
2.6. Impro6ed process stability Process stability is reflected in lower standard deviation of quality variables. McPherson, Crosby, Delaney and Badgwell (1987) reported a reduction of 50% in octane for a catalytic reformer. This APC benefit in fact lies at the heart of all above benefits. In general this benefit is always existent whenever a regulatory or advanced process loop is keeping a production target. A common example of improved stability is the feedforward loop to heaters. In here, this loop will reduce the effects of any fluctuation that takes place in feed coming from upstream unit on a downstream unit. This fluctuation may manifest itself in feed flow rate, temperature, or composition changes. The quantification of the benefits resulting from using any advanced control strategy is the most critical step, which a process engineer takes to decide whether to implement such a strategy. Naturally, if the engineer is considering the application of several strategies, in different parts of the plant, then he has to determine the benefits of all strategies under consideration on performance throughout the plant. The benefit calculation itself is an activity, which is deep-rooted into the
process basic principles, in term of mass and energy balances, coupled with its dynamic properties. Starting from a base case operation of a unit, and its performance objective, the engineer determines if a given advanced control strategy could force the unit to operate around more economical conditions. The APC benefits are thus quantified as the difference of the economic measure of unit operations in the two different operating conditions. The payback period of an APC strategy is then the period in time in which the APC benefits would pay for the cost of APC strategy implementation, which includes, controller analysis, design implementation cost, and the cost of any additional hardware such as process analyzers. As an example for benefit calculation, we consider the case of a residue hydrotreater unit processing the bottom products of a crude distillation unit. The process objective is primarily to reduce sulfur contents of the residue. In addition to the treated residue other products are produced due to some cracking reaction that takes place. These include some light components, naphtha and middle distillates. APC strategies that can be implemented for this unit together with the claimed benefits are listed in Table 1. For each possible APC strategy, we consider the following possible benefits: Increased throughput: if the throughput of the unit is limited by any constraint, and the control strategy could relax this constraint by certain amount, then the capacity credit benefit is the total net economic value of the additional throughput. One such strategy is temperature controller of feed heater. Yield benefit: if the control strategy could alter the yield pattern of different products in such a way that a more valuable product as middle distillate flow is increased by reducing the residue flow, then the benefit is the net change in economic value the flows before and after the change. Such strategies usually
Table 1 APC strategies and claimed benefits for a residue hydrotreater unit APC benefits
APC strategies Multi-variable controller (MVC)
Increased throughput Increased yield Reduced quality giveaway Reduced energy consumption Reduced reprocessing cost Improved process stability
With analyzer
Without analyzer
X
X
Feed-forward control
X
X X
Ratio control
X X
X X
X
X
X
Constraint control
Model based control
X X X
X
X
H.M.S. Lababidi et al. / Computers and Chemical Engineering 26 (2002) 1303–1319
involve moving closer to operating constraints such as cloud point, 50% boiling point and endpoint.
Both throughput and yield benefits are functions of the streams flow rates and values. An increase in throughput or yield would result in altering the material balance of the process. For this reason, the throughput and yield benefits are incorporated in the mathematical model (see Eqs. (4), (5) and (11)). Quality giveaway: for the hydrotreater, quality giveaway is mainly claimed from reducing overdesulfurization. If the reactor’s severity is set in such a way where sulfur reduction is larger than desired (sulfur content of the product stream is lower than required), then the benefit of reduction of desulferization could be computed using the value of reduced hydrogen consumption, increased throughput of the reactor, and possibly longer life of reactor catalyst. Sulfur giveaway is usually expressed as a penalty correction factor for over-conversion, given as dollars per percent of sulfur removal per ton. The cost of such strategy will include using an on-line sulfur analyzer. Other quality benefits accounted for include flash point and boiling point corrections in the fractionator and related equipment. Energy benefits: energy benefits are considerably high for this type of processes. The heavy residue feed needs to be preheated to high temperatures before entering the reactor. Feed temperature is extremely critical in overcoming catalyst deactivation and directly affecting the catalyst life. APC benefits from reduced energy consumption are accounted for through reduced excess oxygen in the furnaces and temperature pass balancing in the heaters, which would translate into higher efficiency, and lower fuel consumption. Reduced reprocessing and improved process stability are considered as intangible benefits, which minimize variations in the product and diminish the risk of degradation or reprocessing out of spec material. If a given controller could reduce the effects of process disturbances, in such a way where the unit is driven away from some constraints toward more economic conditions then the benefits of this strategy would be the net difference in economic value of unit operation. In addition, this may result in more consistent product property, thus precluding the need for reprocessing. Compared with the other benefits, reduced reprocessing and improved stability benefits are usually referred to as less-quantifiable benefits (Goodhart, 1998). The actual quantification of those benefits prior to APC implementation is most of the times difficult. However, an accurate evaluation of outcome process performance with APC may bear out the true value of such benefits. Refinery managers usually keep a worried eye on
1307
their plant profitability. They tend to view APC simply as means to raise their profits. However, this robs APC of their true benefits, where they help deal with specific operational issues that have indirect benefits. A typical situation that applies here is when an upstream unit is continuously disturbed. So by applying APC, to such unit, even when no direct economic benefit is apparent, will lead to improved stability, thus reducing the disturbances impacting downstream units. This is why Goodhart (1998) considered that ‘APC benefits come from smoother operation of plants (reducing impacts of process disturbances) and providing consistent operation at optimal constraints’. The economist uses the term opportunity cost. This may apply very well to APC, where the failure to apply it could translate into higher cost. Thus APC should not be looked as from a narrow benefit angle, but rather from a wider angle reflecting production priorities. In fact, the priority of applying an advanced control strategy in a unit, reflects the importance of achieving the benefits of this strategy in this unit, vis-a`-vis the refinery wide production benefits. For instance, an APC strategy that helps maximize distillate flow rate in a reactor fractionator, while meeting a cold flow property target, could have direct benefits in the gas oil blender down stream, even when the benefits at the fractionator do not warrant the use of this APC strategy. Hence in this example, the use of the APC for the fractionator is decided by APC production priorities, rather than by its direct benefits. The direct implication here is that APC strategies should be used in units, where it can lead to direct benefits, or even without benefits, if it can alleviate upstream or downstream unit constraints. For this reason APC implementation decisions are best made in light of APC benefits and priorities, which are best reflected, in the refinery wide benefits. In the course of this work, a practical example was encountered, where the common operational wisdom indicated that there is no direct benefits to strict cold flow target control at upstream units, since it can be corrected in the product blender. However, it turned out that due to physical and operational considerations means available in the blender led to final products that were more expensive to produce. This practical example confirmed the importance of applying APC strategies at a unit, where deviations from production targets occur. A case in point on the priority of APC strategies is reducing hydrogen consumption. While the monetary benefits are limited to a conversion unit like the hydrocracker, relieving this constraint can enable other units to operate at more desirable severity conditions. So in this case such strategy will take a far higher priority than other strategies that have more exclusively enumerated benefits. In most APC master plans, recommendations on the implementation of APC strategies are based on direct
H.M.S. Lababidi et al. / Computers and Chemical Engineering 26 (2002) 1303–1319
1308
benefits, and shortest payback periods. In here, we are providing a platform where such decisions are made using a criteria far more comprehensive, and more relevant than the payback period. Specifically, we have identified several objective functions that could be used to derive the recommended list of APC strategies. These objective functions reflect direct and indirect benefits. The more important aspect here is that the model can always serve as a sensitivity analysis tool which would tell the decision maker how his recommendations would change if any of the indirect benefits were included in the model. Another, use of this model is to update APC decisions when any of the direct benefits is drastically changed by operational factors, such as change of catalyst, or alteration of product slate profile. The model is integrated with an optimizer to form the advanced control-planning tool. A summary of the capabilities and uses of this tool is given in the sequel.
3. APC planning model The planning model is developed for a set of units, Nu, for which APC strategies are applicable. The units are given a sequential number from i= 1 … n. Candidate APC strategies applicable for unit i are defined as NSi. Binary decision variables, Sik, are defined to indicate the selection of APC strategies. Sik =
!
1 0
if strategy k is selected for unit i otherwise
(1)
Model formulation is described in the following paragraphs. The objective function will be discussed first followed by the constraints.
3.1. Objecti6e function Any refinery would be interested in seeking the maximum monetary return on the investment for introducing APC. This entails developing an objective function, which consider both the net cost reduction (NCR) and the net change in the monetary value of products (MVP). For the NCR criterion, there is a trade-off between the cost of introducing a given APC strategy versus the reduction in unit operating cost to be realized by using this strategy. This optimization criterion can be expressed mathematically as:
NCR = %
% (rik −Cik )Sik
i N u k NSi
(2)
rik is the cost reduction in unit i due to introduction of APC strategy k, and Cik is the implementation cost of this strategy. Cost reductions (rik ) are determined for the six types of benefits discussed above. The introduction of APC strategies could lead to increasing throughput or yield of more valuable products. For instance, the cost reduction of increased yield is quantified by multiplying the increase in yield by the product value. The objective function for maximizing valuable products is then expressed as: MVP = % % (Xij − X( ij )Vij
(3)
i N u j Bi
Xij is the nominal flowrate of the stream flowing from unit i to unit j, while Xij is the actual flowrate after introducing APC, and Vij is the value of this stream. Bi is the set of units upstream of unit i. Costs and benefits are calculated on daily basis, and annualized for economic calculations such as the payback period. The proposed objective function for planning APC strategies is the NMR, which can be represented as the sum of NCR and MVP objective function given by expressions Eqs. (2) and (3). Hence, the problem is to maximize NMR expressed mathematically as: NMR =
!
%
% (rik − Cik )Sik
i N u k NSi
+ % % (Xij − X( ij )Vij
"
(4)
i N u j Bi
This formulation introduces the least sophisticated model for the optimum selection of APC strategies. Several extensions can be introduced into the model in order to realize additional functionality in planning APC strategies. One important extension is minimizing the variation in product quality. The quality of a product can be identified by its properties that are considered critical or limiting. Common refining properties are API, cut point, sulfur content, cold flow properties, viscosity, RVP and octane. Adding such feature to the objective function would result in favoring the APC strategies that cause minimum variation in product specifications. As in the operation of any process unit, operating variables are continuously perturbed by property and flow variations, and load changes. Extent of perturbation is measured by the standard deviation of key process variables. Tight regulation resulting form using APC strategies will invariably lead to considerably lower standard deviation, which in turn will reduce product flow and quality variations. In practice, variability is dealt with operationally at the unit level. However, the present planning model has been developed as a decision tool for determining APC implementation on a refinery-wide considerations. In view of the above points, inclusion of unit related uncertainty
H.M.S. Lababidi et al. / Computers and Chemical Engineering 26 (2002) 1303–1319
would unduly complicate the problem without adding significant impact on the selected APC implementation phase.
1309
Uji is the upper limit on feed flow rate to unit i from unit j.
3.2.2. Incompatibility of APC strategies 3.2. Constraints The objective function represented by Eq. (4) is subject to the following constraints:
3.2.1. Material balance and flow adjustments Adjustment of mass flow rate from unit i to unit j due to APC strategies applicable to unit i is given as: Xij = X( ij + % gijk · Sik +Y( ij % hik · Sik k NSi
k NSi
Öi and jNu
(5)
The above equation reflects the benefit from APC strategies, emanating from yield pattern change, and capacity credit change. gijk is the change in flow from unit i to unit j due to APC strategy k, while hik is the capacity benefit of unit i due to strategy k. Yij is the nominal yield for the stream flowing from unit i to unit j. Although the APC strategies may alter the yield, to maintain linearity, the effect of added capacity is distributed according to the nominal yield, Y( ij, defined later by Eq. (11). The yield and capacity incremental changes, gijk and hik, were developed following a comprehensive review of a refinery base case operation in which process study data for each unit were used to determine the possible benefits related to specific APC strategies. In addition to the flow adjustment constraint, the following balances are required for the interaction between the units. Balance on flow rate between intermediate units: % Xij − % Xji =0
j Ai
j Bi
Öi NI
(6)
NI is the set of intermediate units. Ai and Bi are the set of units downstream and upstream of unit i, respectively. Balance on flow rate out of source units: % Xij = ai + % hik · Sik
j Ai
k NSi
Öi NS
(7)
NS is the set of source (feed) units and ai is the amount of feed available for source unit i. Balance on flow rate to destination units: % Xji =bi
j Bi
ÖiND
(8)
ND is the set of destination (product) units, and bi is the amount of flow required out of destination unit i. Unit capacity limit due to APC strategies of unit i. Xji 5 Uji + % hik · Sik k NSi
Öi Nu
(9)
% Sik 5 1
ÖiNu
k i
(10)
i is the set of mutually exclusive alternative strategies for unit i. Constraint Eq. (10) will be used to eliminate the possibility of duplicating an APC strategy, or when two strategies are similar in function but they provide different benefits. For instance in the case of using an MVC without analyzer, or an MVC with analyzer.
3.2.3. Yield relationships and benefit Input and output streams are related by yield functions. The nominal yield of the stream flowing from unit i to unit k is defined as: Y( ik =
Xik
or
Xik = Y( ik % Xji
% Xji
j Bi
j Bi
Öi and kNu (11)
The total yield benefit due to all strategies in unit i can be then expressed as: BFi = %
% gijk · Vij · Sik
j Ai k NSi
Öi Nu
(12)
Vij is the value of the stream flowing from unit i to unit j. The proposed APC planning problem can be summarized as maximizing the net monetary return (NMR) objective function (Eq. (4)), subject to the material balance adjustment constraints, Eqs. (5)–(9), and the incompatibility constraint, Eq. (10). The APC planning model contains N*NS u i binary (0, 1) variables. Nu is the number of units in the refinery flowsheet considered for APC, and NSi is the number of possible APC Strategies applicable to unit i. The number of continuous variables, Xij will vary depending on the streams connecting pairs of units as well as the number of properties to be considered in the model. One of the most important features of the proposed model is considering capacity credits and the way those credits are reflected on the process flowsheet. Capacity credit multipliers are introduced to indicate an increase in unit throughput due to the application of an APC strategy. The model includes the capacity credits in both, the objective function and the constraints. The objective function (Eq. (4)) considers the value (monetary benefit) of the capacity credits in its second term, whereas the effects of the capacity credit on the material balance are represented in the constraint Eqs. (5), (7) and (9).
1310
H.M.S. Lababidi et al. / Computers and Chemical Engineering 26 (2002) 1303–1319
Fig. 1. Structure of APCO APC planning package.
It is important here to explore and clarify the impact of the capacity credit on the material balance, and the difference between the direct benefits of the APC strategies and the capacity credit benefits. One of the objectives of the APC strategies is to maximize the yields of more valuable products. This involves recovering fractions of streams to more valuable streams. In this case the flow rates of the output streams are altered while the overall throughput or capacity of the unit is not changed. However, the changes in output streams flow rates affect the units down stream, and this effect is included in the material balance Eq. (5). On the other hand, a capacity credit is a claimed increase in unit throughput. According to the mass conservation law, this increase in throughput should be reflected in both the input and output streams of the unit. The importance of adjusting the material balance is in satisfying the capacity limit constraint (Eq. (9)). In other words, if an APC strategy introduces a capacity credit in certain unit, and this increase cannot be handled by a unit further downstream, or supplied by a unit upstream, then this strategy is not applicable. Despite the fact that the capacity credit multipliers are very small (e.g. 0.001%), their effect on the capacity limits might be significant because they are accumulated in the material balance adjustment process.
4. APC planning optimizer (APCO) The APC planning model has been implemented as an APCO package. APCO is an interactive tool for selecting the optimum APC strategies. It enables the user to define the planning scenario, develop the optimum APC plan, view and analyze the optimum plans,
save and compare various planning scenarios and print summaries of planning reports. The APC planning optimizer element of APCO is supported by a database and a graphical user interface (GUI), which facilitate performing the planning studies in a flexible and friendly environment. Main features and functionality of APCO will be highlighted in this section. The design structure of APCO is shown in Fig. 1. In developing APCO we tried to follow the object-oriented paradigm as far as possible. This enabled the program to be constructed gradually from different elements interacting with each other. The database stores the unit operation and the APC strategies data. Benefit and capacity credit calculations are also performed at the database level. A planning session is conducted by interactively specifying the planning parameters, such as budgets, planning period and constraints. The optimizer is then started for selecting the optimum APC plan for the given scenario. The APC planning results can be then visualized at various levels of details, using the GUI. It is also possible to save APC planning results for future reference. APCO has been fully developed using MICROSOFT EXCEL environment. Reasons behind adopting EXCEL for APCO implementation can be summarized as follow: Spreadsheet programs are widely used by the engineering community. Refinery engineers are quite familiar with EXCEL environment. Complexity and dependability of the optimization variables and parameters are ideally handled by spreadsheet calculations. EXCEL worksheets have been utilized as both, a database and an evaluation site for the optimizer. Such dual functionality cannot be found in LP packages. The build in solver of EXCEL was found to be suitable and efficient for solving the APC planning problem. EXCEL is supported by VISUAL BASIC, which is a very rich and flexible programming language. VISUAL BASIC programs were written to automate the APC benefit calculations as well as developing the GUI of APCO.
4.1. APCO database The database consists of spreadsheets representing the operational and economic parameters required for APC planning. Each unit operation in the refinery is associated with a spreadsheet storing the required data. Each spreadsheet consists of three elements. The first one stores the input/output stream flow rates and costs, which are used in calculating the yields and benefits. Stream data is linked with the upstream and downstream data of other units. This facilitates performing
H.M.S. Lababidi et al. / Computers and Chemical Engineering 26 (2002) 1303–1319
complete material balance on the whole refinery. The second element lists all possible APC strategies applicable to the unit, together with their implementation costs. The third element is designated for APC benefit and capacity credits.
4.2. APCO optimizer The optimization model can be considered as mixed integer linear programming (MILP). The discrete variable, binary (0, 1), is Sik which is used to indicate the selection of APC strategies. The model includes also decisions on continuous variables such as throughput, flows and streams properties. Hence the need arises for a MILP solution approach. MILP solution procedures use branch-and-bound with bounds generated by solving Linear Programming sub-problems. MILP solution procedures are available in various mathematical programming software packages such as the mainframe MPSX/MIP as well as PC based packages such as LINDO, CPLEX and GENERAL ALGEBRAIC MODELING SYSTEM (GAMS), among several others (Kallrath, 2000). The APC optimizer element of APCO is based on MS EXCEL Solver, which is an ‘Add-ins’ tool implemented by Frontline Systems (Frontline, 1998). The Solver uses the Generalized Reduced Gradient (GRG2) nonlinear optimization procedure, and simplex method for linear and integer sub-problems, with the branch-and-bound search method. The decision to use EXCEL solver, rather than a conventional MIP package, is supported by the following points: The material balance constraints and the need to introduce capacity credit multipliers added some complexity to the APC optimization model. The amount of data needed to plan APC strategies for the whole refinery is immense. This includes stream flow rates, values and properties. On the other hand, such data is continuously modified and updated. Customization efforts proved to be quite difficult and ineffective. To perform an APC planning session using an MIP package, an interface is needed to prepare the input file from a database, and another one to interpret the output results, and present it in clear descriptive format. In fact, the most difficult customization step was the interpretation of the output results. Hence, the optimization session would be tedious and time consuming, and comparing different planning scenarios would quite difficult. The performance of APCO optimizer has been tested against GAMS (Brooke, Kendrick & Meeraus, 1992). GAMS is a high level language for formulating models with concise algebraic statements. It is independent of the solution algorithms of the solvers, but
1311
passes the problem definition to one of the solver programs. XA solver has been selected (GAMS/XA, 1995). A number of case studies were considered to compare APCO solver with GAMS/XA solver. All case studies showed identical results. APC plans proposed by APCO and GAMS have the same cost, benefit and selected strategies. Such perfect match increases the confidence in APCO solver, and supports the claim that APCO is more convenient and flexible than dedicated MIP solvers in handling the APC planning problem.
4.3. APCO graphical user interface The GUI of APCO consists of a number of windows and menus developed in order to perform the various tasks in an interactive and friendly environment. The main front end of APCO is shown in Fig. 2. It is composed of two sections, the top part presents a summary of the current APC planning problem, while the bottom part constitutes the main control panel for accessing the main functions of APCO. The main functions of the GUI include setting a new APC planning problem, running an optimization session, saving and retrieving previous APC plans and visualizing the planning results. A new APC planning problem is entered interactively by specifying the planning period and the allocated budget. The user is also given the chance to change key parameters such as the benefit and analyzer multipliers, interest rate, payback period and utilities cost. APCO GUI is quite effective and useful in presenting the APC planning results, which can be visualized at different levels of details. Brief summary, comprising the annual benefit, cost and number of selected strategies, is displayed in the main front end as shown in Fig. 2. It is also possible to generate ‘Year Plan’ and ‘Units Plan’ charts which give histograms showing the selected optimum plan for the entire planning horizon and for every year, respectively. APCO provides also two self-contained and easy to use tools. The first one is the ‘APC Planning Results’ tool shown in Fig. 3. This tool is used for analyzing the results and displaying more details about the APC costs and benefits for certain year and/or unit. The second tool is the ‘Flow Sheet’ tool, which is a graphical representation of the refinery flowsheet showing the units considered for APC implementation. A snapshot of the flowsheet window is shown in Fig. 4. The units displayed in the flowsheet window are active graphical objects. Clicking one of the unit objects will display a new window giving more details about the unit. Such functionality provides high flexibility in accessing and navigating unit’s data and planning results. It also provides the ground for customization as well as integrating other refinery packages with APCO.
1312
H.M.S. Lababidi et al. / Computers and Chemical Engineering 26 (2002) 1303–1319
Clicking on the ‘Crude Unit’, for instance, would display the window shown in Fig. 5. This window provides various information related to the selected unit. Other functionality of the GUI includes saving and retrieving APC planning results, and automatic generation of APC planning reports. Generated reports consist of a summary page followed by detailed lists of the strategies selected for every year showing all benefits and costs and supported by charts. It is also possible to customize the generated report and specify the required format.
4.4. APC strategies and benefits The APC planning system was based on actual and projected benefits for a large refinery in Kuwait. However, to demonstrate the main features of the system without disclosing actual refinery data, a simplified flowsheet is used. The proposed refinery flowsheet is oriented basically to produce auto diesel, LPGs, and gasoline (Jones, 1995), and shown diagrammatically in Fig. 6. A single-stage crude column (i.e. no vacuum distillation) produces full-range naphtha cut overhead,
Fig. 2. APCO graphical interface front end.
Fig. 3. APCO Planning results tool.
H.M.S. Lababidi et al. / Computers and Chemical Engineering 26 (2002) 1303–1319
1313
Fig. 4. APCO flowsheet tool.
Fig. 5. APCO units visualization tool.
a small kerosene side-stream, light and heavy gas oil streams, and an atmospheric residue bottoms. Those streams are further processed and treated by unit operations that are common for most refineries. Various types of data were needed for benefit calculations and APC planning. This includes flow rates, specific gravities, stream prices, a number of physical
properties, and temperatures. Required information was extracted partly from Jones (1995), which is the source of the balanced simplified flowsheet (Fig. 6). APC benefits are calculated on unit basis due to the fact that APC by definition is a local function limited in scope to a unit, and linked to process dynamics and disturbances. Refinery wide benefit is considered as the
1314
H.M.S. Lababidi et al. / Computers and Chemical Engineering 26 (2002) 1303–1319
cumulative sum of all unit benefits. Possible APC strategies have been identified for each unit, together with their implementation cost and tangible benefits. Type of control benefits in a unit may take any combination of: increased throughput, increased yield of more valuable products, reduced quality giveaways, reduced energy consumption, reduced reprocessing, and improved process stability. Those benefits may materializes or not in every unit. Yield pattern change of each unit was quantified if it was economically or operationally significant. One example of quantified tangible benefit is increasing a distillate, or kerosene yield, or reduced severity, and in turn reduced hydrogen consumption. APC strategies are generally classified as conventional and multivariable. Conventional strategies consist mostly of single loop controllers where quality measurement is used, such as measuring or inferring Reid Vapor Pressure and adjusting a fractionator overhead. Feedforward loops are also considered advanced control loops, where disturbances are compensated for. A multivariable controller (MVC) is designed to meet some performance objectives to meet certain economical targets for a unit. By way of example we will consider the role of a MVC for the crude unit. The main objective is to improve the yield of more valuable cuts. Additional objectives are met by maintaining product specifications, improving energy recovery, maximizing feed rate and maintaining stable fractionation tower operation. From economic point of view, feed rate maximization has typically the highest economic return. The MVC manipulates the set points of key loop controllers, which are normally adjusted by the operator. These are the column overhead temperature, feed heater outlet temperature, pump-around flows, and total tower feed rate. The controlled variables, such
as product boiling points, are measured online by distillation analyzers or inferred using secondary measurements. Due to the considerable cost of some online analyzers, two MVC strategies have been identified in this study, one without analyzer and another one with analyzer. A benefit multiplier was introduced to reflect higher benefits in case an analyzer is used. Other APC strategies include excess oxygen control, temperature pass balancing control, heat exchanger trains control, and non-linear level control. The benefit of the excess oxygen strategy is manifested in fuel reduction. The temperature pass balancing strategy is aimed at maintaining the individual pass outlet temperature at desired heater outlet temperature. Hall (1985) demonstrated that a pass balancing strategy for an 83 000-bpd vacuum tower resulted in energy savings which exceeded $300 000. The heat exchanger trains controller alters the flow rate through different trains to maximize heat recovery. The main objective of the non-linear level control is to reduce flow disturbances to downstream units. The main concept is to permit the level to swing at desired range, while maintaining the manipulated variable for the loop at relatively stable conditions (Korchinski, 1995). 5. Case studies Several case studies have been designed to test and demonstrate the capabilities of the advanced process control-planning tool, APCO. The case studies were meant to reflect situations encountered by decisionmakers at the refinery. This includes moving APC projects from the conceptualization stage to the final approval stage and then passing through the most critical stage namely budgeting. Key parameters used in developing the case studies are:
Fig. 6. Refinery simplified flowsheet.
H.M.S. Lababidi et al. / Computers and Chemical Engineering 26 (2002) 1303–1319
1315
Fig. 7. APC plan for unlimited budget case. Table 2 APC plans for different payback periods Plans
A1
A2
A3
A4
Payback (year) Selected strategies Cost ($MM) Annual benefit ($MM/year)
1 21 3.597 Base-1 11.885 Base-2
2 34 4.567 Base-1+0.97 12.545 Base-2+0.66
3 41 5.082 Base-1+1.485 12.783 Base-2+0.898
7 42 5.105 Base-1+1.508 12.735 Base-2+0.85
Budget: capital investment allocated for 1 year or several years. Payback period: the planning model selects APC strategies that will pay for itself in a pre-specified payback period. Several payback periods were examined. For self-funded plans, the optimizer determines the payback period. Interest rate: the discount rate used to compute the present worth of APC benefits. Benefit multiplier: setting the benefit multiplier to 1 reflects tangible or direct and intangible or indirect APC benefits, while switching it to zero would reflect only tangible benefits. Analyzer multiplier: throughout this study, it was assumed that the tangible benefits are hundred percent realized when analyzers are used. The analyzer multiplier was used to reduce benefits when no analyzer is used. Setting the analyzer multiplier to 0.25 or 0.15 reduces the APC benefit 25 or 15%, respectively. Strategies: allows exclusion of specified strategies from consideration in a planning session. Initially all possible APC strategies are included and the user is given the choice to eliminate individual (or all) strategies related to certain unit. It was found that through manipulation of these parameters, it is possible to select the APC plan that will most likely meet the planning and implementation constraints at hand. The two most significant parameters to work with are APC budget and payback period.
5.1. One-year APC plans What causes a given strategy to be considered by the planning model is its benefit, which is calculated on annual basis. To identify all strategies that are economically feasible, the model was run using a very large (infinite) budget allocated for a single year. In this case APCO selected the strategies that can pay for themselves in the specified payback period. For this case study a typical discount rate of 7% was used with 3 years payback. The outcome was a 1 year plan in which 41 out of 64 strategies were selected, with an initial cost of $5.081 million and annual benefit of $12.8 million. MVCs were selected for the crude unit, reformer, debutanizer, reformate stabilizer and diesel hydrotreater. APC planning results for this case study are illustrated graphically in Fig. 7, which is generated automatically by APCO. It is clear from the histogram that the crude distillation unit has the highest benefit followed by the Naphtha hydrotreater and the reformer. The user will be able also to analyze the planning results and study the influence of individual strategies on the overall plan. The effect of payback period is studied through four case studies with payback periods of 1-, 2-, 3- and 7-years, respectively. APC planning results for these cases are listed in Table 2. Only 21 strategies were selected for 1-year payback period (A1). These strategies can be considered as the most attractive alterna-
1316
H.M.S. Lababidi et al. / Computers and Chemical Engineering 26 (2002) 1303–1319
tives, because they can pay for themselves in 1 year or less. They included most of the multivariable control, and several advanced regulatory strategies. By increasing the payback period more alternative strategies become feasible. This can be viewed by comparing A1 with A2 and A3. The 2-years case introduced 13 more strategies, while the 3 years case almost doubled the 1-year strategies. However, selecting long payback periods would have minor effect on the optimal plan (as shown by A4). Comparing case A2 with A1, which is considered as the base case, it could be seen that one million additional cost would lead to an increase in annual benefit of only $0.66 million. In other words, 27% increase in initial budget would result in 5.6% increase in annual benefit. Plan A2 should be contrasted with plan A3 where 41% increase in budget gives only 7.7% in annual benefit. Thus selecting long payback period plans introduce strategies that are less beneficial, as measured by its benefit to cost ratio. Two years payback period has been selected for the rest cases. During this period it is possible to maintain the unit under the same operating condition, and hence the same benefits would hold.
5.2. Multi-year APC plans APC in a large refinery is never introduced in 1 year. Rather it is done over several stages, extending over more than 1 year. This capability is built into APCO where it is possible to compute the optimal steps that could be followed to maximize refinery returns, within a pre-specified budget. Fig. 8 presents the results of a 5 year implementation case study, with an annual budget of one million, and using a 2 year payback period and 7% interest rate. Unlike the unlimited budget cases discussed above, multi-year plans are different in the sense that the model is trying to squeeze in the maximum number of strategies within a given year. The optimizer selects the highest benefit application in the
first year, and left the less beneficial ones for the following years. This is shown clearly by the 5-years plan (Fig. 8) where the annual benefit dropped 30% in the second year and 70% in the third year. Another item to look at in the 5 years plan is the number of APC strategies selected for each year. The number varies from 13 strategies for the first year, only two for the second year, 14 for the third year, and dropped down to 3 in the fourth year. This demonstrates that APC strategies are different in both implementation costs and implied benefits. Since multivariable APC strategies are typically leading to high benefits, the system selects the largest number of the MVC strategies within a given year. The remaining amount of the budget for a given year is spent on smaller and less beneficial strategies. The 5 year plan favored three MVC strategies for the first year related to the crude unit, the depropanizer and the de-ethanizer. MVC strategies for the reformer and the debutanizer were postponed to the second year, while those for the diesel hydrotreater and the reformate stabilizer were selected for the third year. MVC strategies exhibit also high implementation costs. For this reason the second year budget was allocated for two MVC strategies and nothing else.
5.3. Self funded APC plans The self-funded capability built into APCO helps to demonstrate how the benefits from APC could be plowed back to introduce additional benefits to the refinery. Self-funded plans are conducted by specifying the first year investment together with the fraction of the benefits to be invested in the following years. Fig. 9 presents four self-funded plans at different re-investment rates, for an initial investment of 1.5 million dollars. By spending 10% of the benefits, it is possible to introduce the remaining feasible strategies for all units in 5 years. This period is reduced to 4 years for 15
Fig. 8. Five-year plan case study.
H.M.S. Lababidi et al. / Computers and Chemical Engineering 26 (2002) 1303–1319
1317
Fig. 9. Self-funded plans.
and 20% rates, and to 3 years for 25% rate. Furthermore, a 5% rate would require 9 years for implementing all feasible strategies. Obviously, the selected strategies for the first year are identical for all plans. Some of the strategies of the second year are common, while those selected for the following years are quite different. This is due to the fact that the most attractive strategies are accumulated mainly in the first year and partly the second year. Other strategies are called on to satisfy the budget constraint, which varies from one plan to another. It is interesting to note that an analyzer was precluded from the MVC strategy of one of the plans due to budget shortage. Annual budgets, which are equivalent to the implementation cost of the strategies, are also shown in Fig. 9. The self-funded capabilities enable the planner to generate and analyze different investment scenarios. It possible, for example, to explore different initial funding alternatives, as well as assigning a different re-investment rate for each year. Such studies might include other parameters such as the payback period, benefit and analyzer multipliers, interest rate, and excluding selected strategies.
5.4. Unit capacity limiting plan One of the main features of the planning methodology presented here is that it includes a mechanism for spotting the effects of real process bottlenecks on any new APC strategies, which have clear capacity credit benefit. APCO allows the user to place constraint on the maximum feed (capacity) of the units. By comparing the resulting plan to a plan without capacity limits, it is possible to get an understanding of how various APC strategies are interacting under actual plant constraints.
In one of the case studies, we applied a bottleneck on the diesel hydrotreating unit. In this case the model did not select a number of strategies for the crude unit in order not to hit the feed constraints on the hydrotreating unit downstream. Strategies excluded include the MVCs, which were assumed to have an indirect benefit of capacity credit. On the other hand, because the products of the hydrotreating unit are sent to blending, strategies selected for other units were not affected. In another case, a throughput constraint was set on the crude distillation column. This was reflected in the catalytic reformer, where any use of APC would have stipulated additional flow to come from the upstream units including the crude unit, which was not possible. Hence no reformer strategies were included in the plan. This demonstrates the strength of the proposed optimization model and the procedure used to reflect the capacity benefits throughout the refinery flowsheet. The material balance is adjusted to account for capacity benefits resulted from applying APC strategies. Giving the optimizer the ability of adding capacity limit constraints on certain units provides an opportunity to investigate wider range of alternatives. One attractive option is checking the possibility of increasing the yield of certain units and their impact on other units. Another possible problem is studying the feasibility of eliminating a capacity limit and whether APC benefits would pay for it.
6. Discussion and conclusions The development of an APC planning methodology has been presented in this paper. The methodology is supported by an interactive planning environment, APCO, capable of generating comprehensive plans for
1318
H.M.S. Lababidi et al. / Computers and Chemical Engineering 26 (2002) 1303–1319
introduction of APC. The system sifts through a large set of alternatives, spanning several years and picks the most attractive implementation option. This is done while factoring into the selection critical parameters related to benefit calculations, effects of analyzers, and different return rates and payback periods. APCO provides a user-friendly graphical interface, which enables the user to easily visualize, analyze and report the selected optimal plans. Given the flexibility built into APCO, it is possible to study the effects of APC strategies having more complex benefit structure than the one reported in this paper for the simplified flowsheet (Fig. 6). As a matter of fact, APCO has been originally developed for a large refinery with a large number of actual refinery APC benefits. Hence, APCO can be considered as domain independent that can be adapted to other applications including refineries, petrochemical or any process industry where APC is being introduced. This can be achieved by altering only the database component of APCO, where the stream data and benefit calculations are specified. The GUI of APCO is quite friendly and flexible. It provides the user with various tools to specify the planning problem, run the optimizer and visualizing the results. The object-oriented approach followed in designing the overall structure of APCO was found very advantageous in extending its features and in adapting the system to another application. All components are considered as independent objects communicating with each other through methods and messages, utilizing a central database. The system includes also material balance and yield modules for all process units as well as a product cost module. These modules can be linked to the refinery plant information system in order to reflect new changes in the refinery into the model. This way it is possible to have an up-to-date online model capable of reflecting the effects of APC implementation on the overall refinery profitability. APCO environment is fully based on MS-EXCEL, which was found very satisfactory. From the end-user point, EXCEL is available just about to any engineer, and most refinery professionals are conversant in spreadsheet programs. On the other hand, the built-in optimization tools and programming capabilities were found very efficient and useful. For the APC planning model, all planning results were found identical to those generated by GAMS, which is a specialized optimization package. Given that budget for new projects is always heavily contested between various modernization, and improvement projects, exercising APCO will enable the decision makers to reach very objective decisions. Moreover, APC planning is not a one-time activity. The reality of operating a refinery, for example, and facing operational and market changes, reveals that APC plans are never static, because conditions leading to their selec-
tion are changing. Hence, it is required to continuously re-evaluate the optimal APC plans and their schedule of implementation. For this reason the planning tool becomes very important for selecting more beneficial strategies and reflective implementation schedule. Another built-in flexibility of APCO is that it allows excluding strategies that were already implemented, or should be excluded for operational reasons. Such capability could be used to reflect actual implementation of APC strategies. Consider for instance a case where a multi-year plan was developed. After actual implementation of the first year strategies, the planner would evaluate the implemented strategies, feedback necessary changes to the benefit calculations, and re-schedule the implementation of the remaining strategies. APCO also facilitates developing and evaluating different economical scenarios. Self-funded plans, for instance, are quite important in justifying initial investments. Planning for different payback periods and return rates is also important in comparing alternative plans.
References Agnihotri, R. B., Bourgeois, L. G., Crosby, J. E., & Hammann, M. D. (1987). Predictive control enhances H- oil hydrocracker operation. Hydrocarbon Processing 66 (6), June, 47 – 49. Bozenhardt, H., & Dybeck, M. (1985). Estimating energy savings for process control computer projects. American Institute of Chemical Engineers, National Meeting. Brooke, A., Kendrick, D., & Meeraus, A. (1992). GAMS-A User’s Guide, release 2.25. The scientific press series. Massachusetts, USA: Boyd and Fraser Publishing Company. Bullerdiek, E. A., & Hobbs, J. W. (1995). Advanced controls pay out in 6 weeks at Texas refinery. Oil Gas —European Magazine, 3, 37 – 41. Cutler, C. R. (1983). Dynamic Matrix Control: an optimal multivariable control algorithm with constraints. Ph.D. Thesis, University of Houston. Cutler, C. R., & Ramaker, B. L. (1979). Dynamic matrix control — a computer control algorithm, American Institute of Chemical Engineers National Meeting, Houston, Texas. Davis, J., James, D., Catt, M., & Adair, D. (1986). Computer control of a desulfurizer fractionator. Chemical Engineering Progress, 82 (3), 62 – 66. Eriksson P., Tomlins A., & Dash, S. K. (1992). FCCU advanced control system achieves 2-month payout. Oil & Gas Journal (OGJ Special) 62 – 78. Friedman, Y. Z. (1994). Model-based control of crude product quality. Hydrocarbon Processing, February, 97 – 106. Frontline Systems Inc. (1998). Premium sol6er plus for microsoft excel, ver. 3.0. Incline Village, NV, http://www.frontsys.com. GAMS/XA (1995). GAMS -The sol6ers manuals. GAMS Development Corporation, USA. Goodhart, S. G. (1998). Advanced process control using DMCplus™. UKACC International Conference on Control ’98, 1 – 4 September, 439 – 441. Grosdidier, P., & Kennedy, P. (1990). Case study in multivariable control-main fractionator slurry pumparound. The Chemical Engineering Journal, 44, 119 – 127.
H.M.S. Lababidi et al. / Computers and Chemical Engineering 26 (2002) 1303–1319 Hall, G. F. (1985). Energy and product savings through advanced control. Hydrocarbon Processing, 64 (7), 57 –59. Humphrey, J. L., & Seibert, A. F. (1992). Separation technologies. An opportunity for energy savings, Chemical Engineering Progress, 88, March, 32 –41. Jones, D. S. J. (1995). Elements of petroleum processing. Baffins Lane, Chichester, England: Wiley. Kallrath, J. (2000). Mixed integer optimization in the chemical process industry, experience, potential and future perspectives. Transactions of Industrial Chemistry and Engineering, 78 (Part A), 809 – 822. Korchinski, W. J. (1995). Advanced level control: explained and analyzed. Hydrocarbon processing —(International ed.), 74, 65 – 66. Korchinski, W. J., Black, D. B., Li, M. C., Warrington, P. S., & Rasmussen, K. (1992). FCCU advanced control in a DCS. Hydrocarbon Processing, 71(11), November, 73 –79. Martin, G. D., Turpin, L. E., & Cline, R. P. (1991). Estimating control function benefits. Hydrocarbon Processing, June, 68 – 73. Mastrogiacomo, M., Bilodeau, D., Treiber, S., & Walker, J. (1997) Advanced control of hydrogen network reduces energy consumption. Oil & Gas Journal, September 1, 72 –75. McPherson M., Crosby, J. E., Delaney, M. C., & Badgwell, T. (1987). Advanced system controls catalytic reformer unit at Texaco Canada. Oil & Gas Journal, (OGJ Report) 33–37. Pelham, R. O., & Moriarty, R. D. (1985.) Survey plants of energy savings. Hydrocarbon Processing —(International ed.), 64, 51 – 56.
1319
Paraschiv, N., Marinoiu, V., Patrascioiu, Cr., & Cartoaje, B. (1996). Advanced control system for crude oil plant — a case study. Computers and Chemical Engineering, 20, nSuppl. Pt 8, S1125 – 1129. Richalet, J., Rault, A., Testud, J. L., & Papon, J. (1978). Model predictive heuristic control: applications to industrial processes. Automatica, 14, 413 – 428. Richard, L., Watson, D., Danzinger, F., Tuppinger, D., Schuster, R., & Wilmsen, W. (1995). Advanced control improves MHC-VGO unit operation. Hydrocarbon Processing 74 (3), March, 163 –172. Rivas, A., & Barsamian, J. A. (1989). Refinery Advanced control and optimization project experience at CEPSA, Proceedings of the ISA International Conference and Exhibition, Philadelphia, PA, USA, ISA Services Inc, Research Triangle Pk, NC, USA, 153 – 160. Takatsu, H., Itoh, T., & Araki, M. (1998). Future needs for the control theory in industries —report and topics of the control technology survey in Japanese industry. Journal of Process and Control, 8 (5 – 6), 369 – 374. van Wijk, R. A., & Pope, M. R. (1992). Advanced process control and on-line optimization in Shell refineries. Computers and Chemical Engineering, 16S, S69– S80. Wellons, M. C., Sapre, A. V., Chang, A. I., & Laird, T. L. (1994). On-line power plant optimization improves Texas refiner’s bottom line. Oil & Gas Journal, 92 (20), May 16, 53 – 58.
