New approach to the characterization of petroleum ... - Semantic Scholar
New Approach to the Characterisation of Petroleum Mixtures Used in the Modelling of Separation Processes
Egon Eckert1 and Tomáš Vaněk2
Department of Chemical Engineering, Institute of Chemical Technology, Prague, Technická 5, 166 28 Prague 6, Czech Republic
Characterisation of complex mixtures is a common tool especially in oil processing industry. Characterisation procedures result in experimentally gained characterisation curves, but for the simulation of industrial processes the definition of a substitute mixture is required. Traditionally, a system of pseudocomponents is derived from the TBP (True Boiling Point) characterisation curve, but there are a number of disadvantages, e.g. the physical properties of pseudocomponents must be estimated by unreliable empirical methods. The new approach to the characterisation of complex mixtures is based on representing the original mixture by a system of real components. Such substitute mixture is fully defined, it has a chemical character, and physical properties can be simply retrieved from databases. Utilisation of a substitute mixture of real components in the simulation of crude oil processing proved that the new approach could replace the traditional one in normal boiling temperature ranges where real components are available. Both approaches could be also easily combined.
Keywords: Refinery processes, Oil processing, Characterisation procedures, Complex mixtures, Pseudocomponents
1
[email protected]; 2 [email protected]
1.
Introduction Mixtures containing an extremely large number of various components can be often
encountered in industrial chemical technologies, particularly in oil processing and refining. Briesen and Marquardt (2003, 2004a, 2004b) analysed thoroughly the past, present and future trends in this branch of industry and pointed out the need to increase the accuracy in the modelling of oil refining processes. There are direct economical and environmental consequences of any progress in this point and improved treatment of complex mixtures is probably the most promising direction. In order to deal with such mixtures in modelling and simulation calculations, it is necessary to simplify the problem by utilising a substitute mixture possessing reasonably lower number of components or to use an alternative representation of the mixture. For this purpose the continuous thermodynamics can be employed (see e.g. Rätzsch & Kehlen, 1983), or, more recently, the wavelet-Galerkin discretization has been proposed (Briesen & Marquardt, 2003, 2004a, 2004b). While continuous thermodynamics has received only a little attention in industrial practice, the adaptation of wavelet-Galerkin discretization method for the modelling of unit operations, where complex mixtures are processed, is intensively being studied. Its main advantage is the possibility to tune the representation of the mixture by means of adaptive control of the problem discretization. On the other hand, this approach uses a non-trivial mathematical background and also the standard models of unit operations must be reformulated in order to incorporate the continuous representations (distributions) of some model variables, e.g. of the vector of composition (Briesen & Marquardt, 2003, 2004a, 2004b). Moreover, physical properties dependent on the composition (e.g. K-values) must be also converted to distribution-based functions. Such methods can be employed only for physical processes as distillation or absorption, but not for processes with chemical reactions. Since the last decade of 20th century, methods for the "reconstruction" of the chemical composition have been developed presuming that some other information was available
beside the usual distillation curve. It might be data from elemental analysis, gas chromatography, mass spectrometry, 1H and 13C NMR analyses, etc., (e.g. Whitson & Brulé, 2000). When using substitute mixtures in steady-state or dynamic simulation, the classical unit operation models can be employed. In fact, in simulation programs the solution engine deals equally with real components as well as with pseudocomponents if properly defined during the data input phase. The number of components used for the substitute mixture is usually up to 102, which is considerably lower than for original complex mixtures, i.e. 104 - 107 or more in case of crude oils. It might be felt that in the age of powerful computing machinery the number of components in chemical engineering calculations could be practically unlimited. Nevertheless, there are at least two good reasons why to keep lower number of components. First, the dimension of unit operation models is dependent on the number of components involved and especially equation-oriented simulators could still run into troubles according to memory requirements and internal limits. Second, it is very hard to analyse the results of simulation run when the number of components is high, e.g. thousands or more, and no new information could be obtained. Probably, there would be subsets of components behaving almost equally, the content of each being on the ppm level. It could be noticed that also for well-defined mixtures in practical calculations it is often desired to decrease the number of components or even pseudocomponents. The method called lumping is based on representing groups of components with close boiling points and/or some other properties by a selected single member component inheriting in the mixture the "weight" of the entire group (Riazi, 2005). Montel & Gouel (1984) suggested to optimise the lumping scheme for the substitution of a group of known components in order to preserve the PVT behaviour. The approach based on pseudocomponents had been developed quite a long time ago (Edmister, 1955; Katz & Brown, 1933) and first was used for flash calculations on an early computer (Hariu & Sage, 1965). As the main advantage it was appreciated that the characterisation procedure was non-iterative. It is still widely accepted as a convenient method in the simulation of separation equipment, but a number of problems arise. Above all:
•
The chemical character of components forming the mixture could also play a role in chemical reactions occurring in the studied processes. For pseudocomponents we can't define any chemical character.
•
A pseudocomponent is primarily defined only by its (pseudo-) boiling point and by some additional parameters, mostly by specific gravity, molecular weight or viscosity. All other physical properties, e.g. needed for simulation calculations, must be estimated. Unfortunately, the reliability of common estimation methods (listed for example by Whitson & Brulé, 2000) for critical properties, the acentric factor, etc. is rather low according to the results of testing published for example by Twu (1984), Lindqvist et al. (1994), Riazi (2005).
•
For pseudocomponents we cannot use group contribution methods (e.g. UNIFAC) requiring information about the molecular structure of compounds in order to estimate some parameters (e.g. binary interaction parameters for vapour - liquid equilibrium).
•
The information about the type of the mixture, e.g. if paraffinic or aromatic compounds are prevailing, or about the type of some of its important components (e.g. polar compounds) couldn't be easily utilised.
•
Arbitrary combinations of pseudocomponents and compounds identified in the original mixture are not supported in commercial simulation programs. That is, it is not possible to place a real component into the middle of the temperature range used for the definition of pseudocomponents without knowing its content.
When modelling distillation and absorption processes for complex hydrocarbon mixtures within simulation programs, the original mixture is traditionally substituted by a mixture comprising two different groups of components. For the light end, real components usually up to C5 are employed if they are present and if their content can be determined. For the remaining part, a system of pseudocomponents is derived from certain global characterisations of the mixture based on its distillation behaviour. This approach is implemented in recent major commercial simulation
programs as ASPEN Plus, HYSYS and PRO/II (ASPEN Plus and HYSYS are registered trademarks of Aspen Technology and PRO/II is a trademark of SimSci-Esscor). Beside these standard features, there are most recent attempts to implement other approaches to oil and gas processing, e.g. available with HYSYS 3.2 onwards as HYSYS Upstream™ Option. The usage of pseudocomponents has after all one important advantage: when defined and equipped by a set of estimated physical (pseudo)properties, pseudocomponents could be in simulation programs treated as any other real components (except in processes with chemical reaction). This is also the main presumption for the alternative way how to establish a feasible substitute mixture as suggested Ba et al. (2003). It has been shown that the substitute mixture composed of suitable real components can be used in simulation calculations instead of an alternative mixture of pseudocomponents with comparable results (Eckert & Vaněk, 2003). The primary intention was to overcome the main disadvantages of pseudocomponents mentioned above, namely the need to estimate their properties. It is possible to retrieve property data from the database, which is used as the source of real components. At the same time, wide variety of thermodynamic models and packages of compatible methods, typically within simulation programs, can be used. There is no need to limit to equation-of-state (EOS) models but other types of models, especially for the description of the phase equilibrium, can be employed, even those requiring some interaction parameters. Moreover, the mixture of real components exhibits a chemical character, which can be utilized when modelling complex reaction schemes. For example, Bělohlav et al. (2005) recently used substitute mixtures of real components for the prediction of yields from pyrolysis reactors. Despite the simplicity of the new approach, it has not been used before, e.g. no attempt to employ real components for this purpose is mentioned even in the most current Riazi's monograph (Riazi, 2005). The principles of the new approach makes it directly usable in simulation programs with current content of their databases of physical properties and for current models of unit operations.
Further in this text, we shall recapitulate the two approaches employing substitute mixtures in order to explain the difference between the appropriate algorithms. The verification of the new approach is then provided on a non-trivial simulation case - crude oil refining.
2.
Characterisation of mixtures using pseudocomponents Normally, two steps are needed to define a system of pseudocomponents for a complex
mixture. First, some standard characterisation curve must be experimentally obtained, preferably the TBP (True Boiling Points) curve, which is directly used in the next step. The TBP curve represents the dependence of temperature measured in the head of a laboratory batch column on mass or volume fraction distilled. The column should have a high number of stages and perform at large reflux ratio (API, 1992). There are other possibilities, e.g. the TBP curve can be substituted by its more accurate chromatographic equivalent SIMDIST or by transformation of curves resulting from some other characterisation procedures (ASTM D86, EFV - Equilibrium Flash Vapour) using empirical correlations (API, 1992). The later case is not too reliable and can lead to a considerable deviation of the calculated TBP curve from the experimental curve (Ba et al., 2003). In the second step the range of boiling points of the TBP curve is cut in order to obtain non-overlapping temperature intervals (Ti, Ti+1), for i=1,...,I, continuously covering the entire temperature range. There are various possibilities how to choose these intervals - some recommendations were published, for example, in Whitson & Brulé (2000). Usually, it is sufficient to use about 15 K for normal boiling points up to 700 K, about 30 K within 700 and 950 K and about 50 K for higher boiling mixtures as it is done in the HYSYS simulation program when using automatic cutting option (see Hyprotech, 1998). Each temperature interval represents one pseudocomponent with normal boiling point given by the arithmetic or, more precisely, the integral mean temperature over the corresponding interval of fraction distilled - see Figure 1. The relevant definitions of arithmetic and integral mean temperatures are given by Eqns (1) and (2) respectively.
Tbi =
Tb (Φ iR ) + Tb (Φ iL ) , 2
1 Tbi = R Φ i − Φ iL
ò
ΦiR
ΦiL
i = 1,..., I
Tb (Φ ) dΦ ,
i = 1, ..., I
(1)
(2)
As also illustrated by Figure 1, the interval of fraction distilled (Φ iR − Φ iL ), i = 1, ..., I determines at
the same time the relative contribution of each pseudocomponent to the mixture, which is used to define its composition. For the consequent usage of the substitute mixture of pseudocomponents it is necessary to supply the same structure of physical properties as for real components. Usually, it is not difficult to measure besides the TBP curve also the density and/or viscosity of individual fractions collected in the laboratory column during the distillation test. Moreover, for each fraction the mean molecular weight can be estimated. Therefore, if we use pseudocomponents to represent these fractions then two to four data items are available for each pseudocomponent: the mean normal boiling point and one to three other properties, which can be used to trigger a series of empirical estimation procedures delivering the remaining set of properties - critical properties, acentric factor etc. (see e.g. Riazi, 2005; Whitson & Brulé, 2000). Unfortunately, we should be aware of low reliability of such methods as it was mentioned in the introductory part of this contribution. It is relatively simple to measure global parameters of a complex mixture, i.e. bulk properties as the density, molecular weight, refraction index or viscosity, and this is often delivered with other measurements. Unfortunately, their direct use for the derivation of a substitute or model mixture is known only for the density where it is possible to construct under certain conditions an approximation of the characterisation curve from the Watson factor (Wauquier, 1995). When the global density of the mixture is known and the Watson factor KW =
( 1.8Tb )1 / 3 S
(3)
can be assumed to be constant then it is possible to get an approximation of the density characterisation curve (Wauquier, 1995). This could be an alternative to density data measured for each fraction during the characterisation experiment.
3.
Characterisation of mixtures using real components
The main disadvantages of the approach based on pseudocomponents can be overcome if we employ real components to form the substitute mixture. Of course, the selection of suitable real components and the derivation of the substitute mixture must follow certain criteria and an appropriate algorithm must be defined. We shall use two consecutive phases, i.e. a set of components to be present in the substitute mixture must be first selected and then the composition of their mixture will be adjusted. The prerequisites for the algorithm are as follows:
a. The TBP curve measured for the original complex mixture is available Tb = Tb (Φ )
(4)
where Tb is the measured "True Boiling Point" temperature and Φ is the mass or volume fraction distilled. b. Some other characterisation curve is available, e.g. all or some of dependencies
M = M (Φ )
(5)
ρ = ρ (Φ )
(6)
η = η (Φ )
(7)
where M is the molecular weight, ρ the liquid density (eventually specific or API gravity) and
η is the liquid viscosity. The liquid density curve can be also obtained approximately using the Watson factor as mentioned above. Instead of using Eqns (5)-(7) directly, it is better to convert them into "phase portraits" by eliminating the mass or volume fraction distilled and establishing direct relation between these properties and the boiling point temperature:
M = M (Tb )
(8)
ρ = ρ (Tb )
(9)
η = η (Tb )
(10)
c. The overall temperature range is divided (cut) into a system of non-overlapping temperature intervals continuously covering the range. There are several possible approaches similarly to the definition of pseudocomponents - an equidistant grid, adaptive grid taking into account the shape of the curve or a grid based on boiling points of homological series of alkane compounds - see, e.g., Table 5.2 in Whitson & Brulé (2000). d. A sufficiently large database of chemical components and their physical properties is available. Detailed requirements will be discussed further.
3.1
Selection of components The basic assumption is that a suitable database is available. The "quality" of such database can
be expressed not only by the reliability of data included but also by the extent of the database. Here we take primarily into account the number of components but on its own this is not sufficient. It is important to have satisfactory representation for important families of chemical compounds, in our case especially hydrocarbons or detailed groups as e.g. parrafines or aromatics. At the same time, the range of normal boiling points for these compounds must be wide and the normal boiling points should cover the range uniformly and densely. This is, of course, an ideal situation, which is more or less fulfilled by databases from various sources. In "open" literature we can find, for example, the API database (API, 1992) with 478 hydrocarbons up to C30, normal boiling points up to 720 K and molecular weights up to 420 kg/kmol. Generally, normal boiling points of higher hydrocarbons are very difficult to obtain as the thermal decomposition takes place (Kopsch, 1995). Accordingly, data for higher hydrocarbons (C18+) in databases are frequently estimated or their origin is ambiguous. Nevertheless, their long-term usage in many chemical engineering calculations
giving realistic results gives certain guaranty that also the substitute mixture of real components selected from such databases can sufficiently model the original complex mixture. The internal database of the HYSYS simulation program (Hyprotech, 1998), which we used for test calculations, incorporates 521 hydrocarbons and covers approximately the same range of normal boiling points and molecular weights as the API database. The source of data is not explicitly specified and probably the same limitations can be expected as for API data. Thermal decomposition is also one of the reasons, why the number of available hydrocarbon compounds rapidly decreases with increasing normal boiling point as demonstrated by Figures 2 and 3. This situation leads to a poor chance to find a suitable real component to represent higher boiling point temperature ranges. Unfortunately, the mixtures in oil processing contain a significant portion of higher boiling compounds forming the so-called "heavy end". Some components can exhibit normal boiling point higher than 1000 K and molecular weight exceeding 800 kg/kmol. Normal boiling points and densities, as the minimum information needed, can be found, e.g., in Beilstien database (MDL Information systems, 2005), for about 1850 hydrocarbons up to C42, where only two components are available - 1-cyclohexyl-hexatriacontane a 1-phenylhexatriacontane. Figures 2 and 3 show another interesting feature. With increasing temperatures the range of molecular weights of components having approximately the same boiling point is getting wider, see for example the region around 700 K in Figure 3. There are a number of other databases, which could be potentially used for the same purpose, e.g. DIPPR (Design Institute for Physical Properties at Brigham Young University, 2000) or the suite of internal databases attached to the ASPEN Plus. The reasons why we did not employ these databases are technical as the program providing the selection of components needs a self-standing "flat" data file. Our approach, initially designed for exclusive usage of real components in the substitute mixture, can be extended to combine real components for lower and moderate boiling point temperatures with pseudocomponents for the heavy end. Nevertheless, this demonstrates the potential of the new approach.
The procedure for the selection of real components for the substitute mixture can be made flexible in order to exploit all information about the original mixture. As stated above, if the chromatographic analysis of the mixture is available it could be also a basis for the formation of a substitute mixture. Certain components can be clearly identified and directly added to the list of components in the substitute mixture. If no chromatographic analysis is available but the TBP curve is known, all components for the substitute mixtures can be selected according to the following algorithm:
Step 1: For each primary temperature interval (prerequisite c.) a set of candidate components is
selected from the database. The criterion is that each component selected must have its normal boiling point within the considered temperature interval. If needed, filtration conditions can be set, e.g. reflecting the requirements to include only some families of components or, on contrary, to exclude another components or families. For example, it is known that no olefin components are present in crude oils (Petrov, 1987). On the other hand, at least one component should be available for each interval. If not, then either a set of wider intervals must be used or the interval can be represented by a pseudocomponent. The combination of real components and pseudocomponents is necessary for higher boiling mixtures. Step 2: Exactly one component is selected for each primary temperature interval from the set
of candidate components by comparing their physical properties. We know the normal boiling point of each candidate component and from phase portraits (8)-(10) the "desired" values can be derived and compared with values retrieved from the database. The simplest way how to combine the deviations for different properties is to use a weighted sum of relative differences. The criterion for the selection is then defined by K
å k =1
wk
ζ r ,k ,c − ζ m ,k ,c ζ m ,k ,c
→ min c
(11)
where K is the number of measured properties and ζ m ,k ,c , ζ r ,k ,c are the measured and from database obtained (calculated or simply retrieved) values of property ζ respectively. The expression is calculated for each candidate component, c = 1,...,Ci and the component with the lowest value of criterion (11) is chosen to represent the interval. There are degrees of freedom in the choice of weight factors wk , which can reflect the precision of measurements or some other demands. It should be noted that measured viscosity and density in fact represent properties of discrete mixtures (fractions), which can be hardly derived from properties of contained pure components according to non-trivial and unreliable mixing rules. Sometimes no other curves than the TBP curve are available. Step 2 of the algorithm can be then replaced by choosing a component with its normal boiling point being closest to the mean temperature of the appropriate temperature interval. This is an emergency solution but it can be expected that the selected real component would be close to a pseudocomponent defined traditionally while all properties of the real component are directly available from the database. The result of this phase of the algorithm is simply a list of real components together with their normal boiling points and other properties. If it is desired for some reason to include other components into the substitute mixture, it can be done now. Such an obvious reason is, for example, the confirmed presence of a particular component or component type in the original mixture. Especially polar compounds are in the focus since they strongly affect the phase equilibrium in multiphase systems. A compound can be added to the substitute mixture with or without information about its amount in the original mixture. Both possibilities are aided in the second phase of the algorithm.
3.2
Determination of the composition of a substitute mixture The problem is that the normal boiling point of each selected component can fall anywhere in
the primarily defined temperature interval (Ti, Ti+1), see Figure 4, and no longer it is a mean temperature of this interval. If a model of the experimental characterisation procedure were
available then we could abandon the principle of mean temperatures and this step would consist in a repeated evaluation of the characterisation curve from the model for varying composition of the mixture until a satisfactory match with the experimental characterisation curve is reached. This kind of an optimisation can be easily employed, for example, for the EFV curve (Eckert, 1999). When the TBP curve is available then the procedure suggested by Ba et al. (2003) can be used. Figure 5 illustrates the principles of the procedure, which at first glance could be found similar to the method used to derive the composition of a mixture of pseudocomponents. The presumption that each component should represent an interval of fraction distilled, where its normal boiling point is the mean temperature according to Eqns (1) or (2), is preserved. The difference is that we cannot expect that the intervals could be generally chosen to cover the whole range without gaps and/or overlapping. This is automatically fulfilled for pseudocomponents where the procedure starts from the other end - the mean temperatures are calculated for temperature intervals initially chosen to continuously cover the entire temperature range without overlapping. Nevertheless, in the case of real components the idea is to distribute the intervals to make them cover the range as much as possible with minimum gaps and overlapping. This is an optimisation task where the objective function can be defined as follows:
( ) å (Φ
F ΦL =
I +1
i = LE +1
R i −1
− Φ iL
)
2
→ min
(12)
For each interval the starting and end fractions distilled are denoted by the superscript L (= Left) and R (= Right). Generally, we can utilise the information about light-end components, which can be usually identified in the original mixture together with their composition, and incorporate them into the substitute mixture. The index of the last light-end component is therefore denoted by LE R and the corresponding fraction distilled by Φ LE . On the other end of the range, I +1 can be denoted
by HE (heavy-end), even if no heavy-end components were present and by definition L L L Φ HE = Φ IL+1 = 1 . The reason why only the elements of the vector Φ L = {Φ LE +1 ,...,Φ I } are considered
as independent optimisation variables in (12) is that Eqns (1) and (2) can be treated as implicit
( )
equations for Φ i , i = LE + 1,.., I , i.e. Φi = Φi Φi . Alternatively, the reversed relation R
R
R
L
( )
Φ i L = Φ i L Φ i R could have been used with Φ i R as the set of optimisation variables. It is reasonable to define bounds for the values of Φ i in order to avoid getting some unreal solutions and the L
following set of constraints was considered:
Φ iL ≤ Φ iL+1 , i = LE + 1,..., I
(13)
Φ i−1 ≤ Φ ≤ Φ i+1 , i = LE + 1,..., I L i
Optimisation problem (12) together with constraints (13) can be solved using some standard package, e.g. MINOS (Murtagh & Saunders, 1995) as in our case. In order to obtain a consistent composition of the substitute mixture it is necessary to convert
(
the intervals Φ i ,Φ i L
R
) resulting from the optimisation to mass or volume fractions. In ideal case,
the intervals would exactly cover the entire range, the value of the object function (12) would be zero and the following condition would be fulfilled:
å (Φ I
i = LE +1
R i
)
L R − ΦiL = Φ HE −Φ LE
(14)
Then the length of each interval can be directly taken as the appropriate mass or volume fraction. Nevertheless, this is not usually true according to possible overlapping of the intervals and existence of areas not covered with any interval. The simplest way how to get a consistent
(
composition derived from the lengths of intervals Φ i ,Φ i L
R
) is to "normalise" the vector of mass or
volume fractions:
(
x j = Φ −Φ R j
L j
)(
L Φ HE
) å (Φ I
R − Φ LE
R i
)
− Φ iL ,
j = LE + 1,..., I
(15)
i = LE +1
A similar procedure can be used when some of the real components were identified in the original mixture and their relative amounts, i.e. mass or volume fractions, are known. The second phase of the algorithm was implemented as self-standing program incorporating the MINOS (Murtagh & Saunders, 1995) optimisation package.
4.
Application of the method
In order to approve the suggested approach it was reasonable to simulate real industrial processes employing substitute mixtures. Unfortunately, the data measured, for example, on distillation columns and published in open literature are almost always for binary mixtures and small-scale columns. Data from major distillation columns in crude oil processing are very rarely available. Therefore, an alternative way is to compare at least the two different approaches to substitute mixtures on the same simulation problem, once using a substitute mixture composed of pseudocomponents and then a mixture of real components. An important point is to define some methodology how to compare different substitute mixtures. In a former contribution (Eckert & Vaněk, 2003) we used the concept of a "theoretical TBP curve" as curve of a staircase shape reflecting the distribution of normal boiling points of individual components from the mixture. Such curve we could get in a batch distillation column with the number of stages approaching infinity and very high reflux ratio. In its discrete version the j =i −1
theoretical TBP curve is composed of points with co-ordinates (
åx
j
+ xi / 2 ; Tbi) for i = 1,...,I,
j =1
taking into account for simplicity the arithmetic mean values defined by Eqn. (1). Analogical theoretical curves can be constructed for molecular weight and density, eventually API gravity or some other properties. Theoretical TBP curves were constructed for examples of separations in a distillation and absorption columns (Eckert & Vaněk, 2003) and the match between results obtained by both approaches was excellent, thus showing that substitute mixtures of real components can replace mixtures of pseudocomponents with all the extra benefits, which this approach brings. For these examples only the measured TBP curves were available as input information. The example presented further is based on the "Refining Tutorial" case delivered with the simulation program HYSYS.Plant version 2.1 (abrev. HYSYS) and at the same time this program
has been used for all calculations. The advantage is that HYSYS incorporates a relatively extensive database of hydrocarbons and it also enables to define pseudocomponents in a built-in tool called "Oil Environment".
Example (Hyprotech, 1998) Crude oil is processed in a fractionation facility to produce naphtha, kerosene, diesel, atmospheric gas oil and atmospheric residue products. The crude oil is characterised by laboratory assay data in Tables 1.-3. The TBP curve is accompanied by the dependence of molecular weight on the liquid volume fraction distilled and a separate assay on API gravity is also available. A light end is considered and its composition is known. The main flowsheet for the process in consideration is shown in Figure 6. Preheated crude is initially fed to a pre-flash drum to separate vapours from the liquids, which are further heated in a furnace. The pre-flash vapours are re-combined again with the hot crude from the furnace and the resulting stream is then fed to the atmospheric crude column for fractionation. Detailed description and specifications are presented in the Appendix. The icon of the distillation column in Figure 6 represents in fact a complex separation process, which can be in the HYSYS program further represented by an embedded flowsheet, i.e. subflowsheet in HYSYS terminology. The appropriate subflowsheet for our example is shown in Figure 7. The PengRobinson property package was used in HYSYS for the calculation of physical properties. In the "Oil Environment" the standard setting of estimation methods was used, i.e. Twu critical property correlation for molecular weight (Twu, 1984), constant Watson factor method for specific gravity and Lee-Kesler method for critical properties, the acentric factor and ideal enthalpy (Kesler & Lee, 1976). It was intended to solve the simulation case using both methods for the characterisation of the process feed (Preheat Crude stream). Particularly, Table 4 globally summarises how the substitute mixtures were assembled. Temperature ranges defined in the first column of the table were for pseudocomponents divided uniformly to intervals of the same width. For real components this was
also done but, as we know from the description of the algorithm above, it serves only for the initial selection of subsets of real components as candidates for representing these intervals. For both substitute mixtures the same light-end (Table 1) was supposed. Unfortunately, higher temperature ranges had to be covered by 10 pseudocomponents, as there were only few hydrocarbons in the HYSYS database with boiling points exceeding 426.7 oC. Our program used for the selection of real components, accepts input data in various forms and physical units and provides necessary conversions. For our example the conversions of characterisation curves from tables 2 and 3 to phase portraits expressed by Eqns (8) and (9) were done using a piece-wise linear interpolation. The phase portrait for the API gravity has been extended to include the light end components. The reason is that this curve bends at the boundary between light end and the remaining part of the mixture, which affects interpolation or extrapolation operations needed in the algorithm. Then for each temperature interval a set of candidate components was chosen from the HYSYS database having normal boiling points within this interval. This primary selection was limited to hydrocarbon compounds only but also compounds known to be missing in oil fractions (e.g. olefins, cycloolefins) were excluded. Technically, this is enabled by registering the chemical family of compounds and setting an "oil" flag on or off for each component in the database. The most suitable component for each interval was found according to criterion (11) for K = 2, i.e. taking into account the molecular weight and API gravity. The weight factors given to the contributions of both properties were simply set to be one. The resulting list of real components for the substitute mixture is presented in Table 5 and the comparison of the measured and retrieved values of both considered properties are depicted in Figures 8 and 9 as phase portraits. Table 5 contains also the information about the number of candidate components for each primary temperature interval. It is apparent that their number rapidly decreases when approaching higher temperatures. It is also interesting that starting from interval number 16 almost all components selected belong to a monotone series of alkanes. Only components number 18 (n-hexylbenzene) and 23 (n-decylbenzene) do not follow this fact and show
larger deviation from the phase portrait as documented by Figures 8 and 9. Another component showing larger deviation in Figure 9 is n-butylcyclohexane in interval 15. In order to improve the selection it could be possible, for example, to try larger temperature intervals or their different distribution, but it would disable to compare the new approach with the traditional one. Therefore, the actual result of selection was left unchanged and passed to the second phase of the algorithm. After selecting a unique component for each primary temperature interval the actual intervals
(Φ
L i
,Φ i
R
) attached to each component were computed in the second phase of the algorithm and the
composition was derived using Eqn. (15). Table 6 contains the complete overview of resulting substitute mixtures including the light and heavy ends. Pseudocomponents were generated in HYSYS Oil Environment and their names chosen by HYSYS reflect the mean temperatures of appropriate intervals (in deg F). For the new approach the interpretation according to HYSYS could be that the four light-end components (originally 1.13 vol. %) together with the system of selected real components would form an extended light-end (in our example it will comprise 32 components and 63 vol. % of the mixture). The measured and theoretical TBP curves of the Preheat Crude are compared in Figure 10. The comparison of experimental additional curves with properties of real components selected for substitute mixture is depicted in Figures 11 and 12. In these Figures it can be also observed that in the region of higher boiling temperatures program HYSYS is not so successful in the choice of pseudocomponents and a considerable deviation between the experimental and estimated values of the molecular weight as well as API gravity is apparent. The simulation calculations of the entire crude oil fractionation process employing both approaches to the characterisation of complex mixtures run without any problems. The excellent match between both approaches can be demonstrated by Figure 13, where the comparison is provided for all the products of fractionation, which have distinct mean boiling points. The plotted theoretical TBP curves were recalculated in order to eliminate water. In fact, the TBP curves reflect the match in the composition of product streams, as the most important parameter, after processing of a single feed stream in a number of unit operations within a relatively complex technology with
several recycle streams. Also for other parameters of the product streams, as flowrates, temperatures and enthalpies, the results of simulation proved very nice match. It can be noted that these results were reached despite the fact that the database used is relatively small compared to other data sources. Moreover, the content of some of the real components selected into the substitute mixture resulting from the second phase of the algorithm is practically negligible (see Table 6 - components number 6, 14, 18, 28, 30).
5.
Conclusions
The results presented in this paper allow us to say that the new simple approach to the characterisation of complex mixtures is fully acceptable even for simulation calculations of largescale and complex processes including various mass and heat transfer operations. It can replace the traditional approach based on the definition of pseudocomponents for low and moderate normal boiling points. Of course, there is an important assumption about the availability and sufficient number of real components with normal boiling points in the considered temperature range in the database. If necessary, we can always use a combined approach, as in the example above, adding pseudocomponents for higher boiling temperature range where no real components are available in the database. We can remark that the number of real components not only in the HYSYS but also in other databases we have been dealing with is still extremely low compared to the possible number of all hydrocarbons. It is an interesting fact that considering hydrocarbon molecules with 25 carbon atoms, the number of possible configurations already exceeds 600 millions (Krambeck, 1991) and crude oil may contain a huge number of distinct molecular species in the order 106 (Altgelt & Boduszynski, 1994). Nevertheless, substitute mixtures allow to deal with complex mixtures in modelling and simulation of technological processes without need to dispose with data for such extent of existing compounds. Systematic addition of missing data and inclusion of new components into databases is continuously provided by the vendors of the main databases of physical properties (e.g. Beilstein) and certainly it brings better possibilities for the selection of real
components into a substitute mixture, but even current pallet of available compounds gives good results. This was proved not only by the example presented above but also by some other (Ba et al, 2003; Bělohlav et al., 2005; Eckert & Vaněk, 2003, 2005). There are many benefits when using real components for substitute mixtures instead of pseudocomponents: •
The usage of substitute mixtures with real components can be extended from usual calculations of separation processes to processes with chemical reactions (Bělohlav et al., 2005) as the substitute mixture receives a chemical nature, even if only a "substitute" one.
•
Empirical estimation methods for physical properties are generally not needed. The values retrieved from database are more reliable and precise despite the fact that for higher boiling compounds (approximately C20 and higher) they are predicted only. On the other hand, the knowledge of the molecular structure allows us to use group contribution methods to estimate various physical properties, e.g. the phase equilibrium behaviour of the substitute mixture (and therefore of the original mixture as well).
•
The selection of components can be affected by partial information about components positively occurring in the mixture or about the overall character of the mixture. Particularly, we can recognise in some mixtures components containing nitrogen or sulphur, but there is certain problem with availability of data. For example, the HYSYS database contains about 50 sulphuric compounds the presence of which in oil could be expected. On the other hand, the inclusion of such components into the resulting substitute mixture is very important when modelling separation processes having strong impact on the environment (Eckert & Vaněk, 2005).
The new approach could be also very convenient for the implementation into standard commercial simulation programs where the combination with current proprietary databases and libraries of numerical methods, especially optimisation procedures for the second phase of the
algorithm, could be very profitable. There are also certain possible improvements in the construction of the substitute mixture, e,g. the selection of components an/or the composition can be optimised in order to get best match between experimental characterisation data and results of the modelling of the appropriate characterisation procedure.
Acknowledgement
Authors appreciate the support of the fund MSM 6046137306.
List of symbols
C
number of candidate components in the primary temperature interval
F
objective function
I
total number of real components
K
number of measured properties
KW
Watson factor
M
molecular weight
S
standard specific gravity
T
temperature
x
volume or mass fraction
w
weight in criterion (11)
η
viscosity
Φ
volume or mass fraction distilled
ρ
density
ζ
symbol for property
Subscripts b
at boiling point
c
index of a candidate component
HE
index of the first heavy-end component
i
index of a component or temperature interval
j
index of a component
k
index of a property in criterion (11)
LE
index of the last light-end property
m
measured value
r
value calculated or retrieved from the database
Superscripts L
left edge of an interval
R
right edge of an interval mean value
Literature Cited
Altgelt K., & Boduszynski M. (1994). Composition and Analysis of Heavy Petroleum Fractions. New York: Dekker. American Petroleum Institute (API) (1992). Technical Data Book - Petroleum Refining. 5th ed. Washington: API. Ba A., Eckert E., & Vaněk T. (2003). Procedures for the selection of real components to characterize petroleum mixtures. Chem. Pap., 57, 53. (full text available at http://www.vscht.cz/uchi/procesy/) Bělohlav Z., Zámostný P., Herink T., Eckert E., & Vaněk T. (2005). A Novel Approach for the Prediction of Hydrocarbon Thermal Cracking Products Yields from the Substitute Feedstock Composition. Accepted for publication in Chem. Eng. Technol. (full text available at http://www.vscht.cz/uchi/procesy/)
Briesen H., & Marquardt W. (2003). An adaptive multigrid method for steady-state simulation of petroleum mixture separation processes. Ind. Eng. Chem. Res., 42, 2334. Briesen H., & Marquardt W. (2004a). New approach to refinery process simulation with adaptive composition representation. AIChE J., 50, 633. Briesen H., & Marquardt W. (2004b). Adaptive multigrid solution strategy for the dynamic simulation of petroleum mixture processes. Comp. & Chem. Engng., 50, 633. Design Institute for Physical Properties (2000). DIPPRâ 801. Brigham Young University, U.S.A. Available at http://dippr.byu.edu/product.asp. Eckert E. (1999). Non-traditional Characterization of Petroleum Mixtures in Terms of Selected Components. Collect. Czech. Chem. Commun., 64, 571. (full text available at http://www.vscht.cz/uchi/procesy/) Eckert E., & Vaněk T. (2003). Simulation of separation columns using substitute mixtures. Proc. of the 30th Int.Conf. of SSCHE on CD-ROM. Tatranské Matliare, Slovakia, May 26-30, 2003. (full text available at http://www.vscht.cz/uchi/procesy/) Eckert E., & Vaněk T. (2005). Extended utilisation of the characterisation of petroleum mixtures based on real components. Proc. of the 32th Int.Conf. of SSCHE on CDROM. Tatranské Matliare, Slovakia, May 23-27, 2005. (full text available at http://www.vscht.cz/uchi/procesy/) Edmister W. (1955). Improved integral technique for petroleum distillation calculations. Ind. Eng. Chem., 47(9), 1685. Hariu, O. H., & Sage, R. C. (1969). Crude Split Figured by Computer. Hydrocarbon Processing, 48(4), 143. Hyprotech Ltd. (1998). HYSYS.Plant 2.1 Documentation. Calgary: Hyprotech.
Katz D., & Brown G. (1933). Vapor pressure and vaporization of petroleum fractions. Ind. Eng. Chem., 25(12), 1373. Kesler M.G. & Lee B.I. (1976). Improved prediction of enthalpy of fractions. Hydrocarbon Processing, 55, 153. Kopsch H. (1995). Thermal Methods in Petroleum Analysis. Weinheim: VCH. Krambeck F. (1991). Continuous mixtures in fluid catalytic cracking and extensions, in Sapre A. and Krambeck F. (Eds.), Chemical reactions in complex mixtures. Pages 42-59. New York: Van Nostrand Reinhold. Lindqvist P., Markkanen V., & Happonen V.M. (1994). Simulation of a heavy residue vacuum column. ASPENWORLD '94. November 6-9. Boston, Massachusetts, 1994. MDL Information Systems (2005). Crossfire Beilstein Database. San Leandro, CA: Elsevier MDL. Available at http://www.mdl.com/products/knowledge/crossfire_beilstein/. Montel, F., & Gouel P.L. (1984). A new Lumping Scheme of Analytical Data for Compositional Studies. Presented at the 59th Annual Technical Conference and Exhibition, paper SPE 13119. Houston, Sept. 16-19,1984. Murtagh, B.A., & Saunders, M.A. (1995). MINOS 5.4 User's Guide. TR SOL 83-20R. Department of Operations Research, Stanford University. Stanford CA. Petrov A.A. (1987). Petroleum Hydrocarbons. Berlin, Heidelberg: Springer-Verlag. Rätzsch M., & Kehlen H. (1983). Continuous thermodynamics of complex mixtures. Fluid Phase Equilibria,14, 225. Riazi, M. R. (2005): Characterization and Properties of Petroleum Fractions. Barr Harbor: ASTM. Twu, C.H. (1984). An Internally Consistent Correlation for Predicting the Critical Properties and Molecular Weights of Petroleum and Coal-Tar Liquids. Fluid Phase Equilibria 16(2), 137.
Wauquier, J.-P. (ed.) (1995). Petroleum Refining, Vol. 1: Crude Oil. Petroleum Products. Process Flowsheets. Paris: Éditions Technip. Whitson, C. H., & Brulé, M. R. (2000). Phase Behavior. SPE Monograph Series. Richardson: Society of Petroleum Engineers, Inc.
Tables
Table 1. Light end components and their liquid volume %.
No.
Component
Liq. vol. %
1
i-butane
0.19
2
n-butane
0.11
3
i-pentane
0.37
4
n-pentane
0.46
Table 2. TBP Distillation Assay.
o
Molecular weight,
Liq. vol. % Temperature, C kg/kmol 0
26.7
68
10
123.9
119
20
176.1
150
30
221.1
182
40
275.0
225
50
335.0
282
60
399.4
350
70
490.6
456
80
590.6
585
90
691.7
713
98
765.6
838
Table 3. API Gravity Assay.
Liq. vol. %
API gravity
13
63.28
33
54.86
57
45.91
74
38.21
91
26.01
Table 4. Characterisation of the process feed (Preheat Crude), basic temperature intervals.
Temperature Traditional method of characterisation New method of characterisation
range, oC 37.8 - 426.7
28 pseudo-components, uniformly
28 real components
426.7 - 648.9
8 pseudo-components, uniformly
8 pseudo-components, uniformly
648.9 - 760
2 pseudo-components, uniformly
2 pseudo-components, uniformly
Table 5. Resulting selection of real components (without the known light-end).
Number of No. Temp. interval [K]
Component
Tb
M
API
candidate
[K]
[kg/kmol]
grav.
components in interval
5
310.95 - 324.84
2,2-dimethylbutane
322.88
86.18
84.90
3
6
324.84 - 338.73
2-methylpentane
333.41
86.18
83.60
3
7
338.73 - 352.62
n-hexane
341.88
86.18
81.60
3
8
352.62 - 366.51
2-methylhexane
363.20
100.21
75.70
12
9
366.51 - 380.40
2,2-dimethylhexane
379.99
114.23
70.70
9
10
380.40 - 394.29
2-methylheptane
390.80
114.23
70.00
27
11
394.29 - 408.17
2,4-dimethylheptane
406.05
128.26
65.20
23
12
408.17 - 422.06
2,3-dimethylheptane
413.65
128.26
62.30
46
13
422.06 - 435.95
2,6-dimethyloctane
433.56
142.29
61.80
62
14
435.95 - 449.84
5-methylnonane
438.26
142.29
60.60
45
15
449.84 - 463.73
n-butylcyclohexane
454.13
140.27
44.70
25
16
463.73 - 477.62
n-undecane
469.04
156.31
58.60
12
17
477.62 - 491.51
n-dodecane
489.43
170.34
56.50
8
18
491.51 - 505.40
n-hexylbenzene
499.30
162.27
32.58
2
19
505.40 - 519.29
n-tridecane
508.58
184.37
54.60
8
20
519.29 - 533.18
n-tetradecane
526.66
198.38
53.60
6
21
533.18 - 547.07
n-pentadecane
543.77
212.41
51.80
14
22
547.07 - 560.96
n-hexadecane
559.94
226.43
50.60
7
23
560.96 - 574.85
n-decylbenzene
571.10
218.37
33.12
6
24
574.85 - 588.74
n-heptadecane
575.30
240.46
49.50
4
25
588.74 - 602.63
n-octadecane
589.86
254.48
48.60
4
26
602.62 - 616.51
n-nonadecane
603.80
268.51
47.80
8
27
616.51 - 630.40
n-uneicosane
629.65
296.56
46.29
5
28
630.40 - 644.29
n-dodecosane
641.76
310.59
45.74
3
29
644.29 - 658.18
n-tricosane
653.37
324.61
45.04
6
30
658.18 - 672.07
n-tetracosane
664.43
338.64
44.72
3
31
672.07 - 685.96
n-pentacosane
675.04
352.67
44.27
2
32
685.96 - 699.85
n-heptacosane
695.26
380.72
43.46
1
Table 6. Resulting substitute mixtures.
No.
Substitute mixture of pseudo-
Substitute mixture of real
components
components and pseudo-components
Real component /
Liq.vol. %
pseudo-component
Real component /
Liq.vol. %
pseudo-component
1
i-butane
0.19
i-butane
0.19
2
n-butane
0.11
n-butane
0.11
3
i-pentane
0.37
i-pentane
0.37
4
n-pentane
0.46
n-pentane
0.46
5
NBP_109
1.05
2,2-dimethylbutane
2.47
6
NBP_135
0.99
2-methylpentane
0.00
7
NBP_161
1.25
n-hexane
1.54
8
NBP_185
1.50
2-methylhexane
3.04
9
NBP_210
1.64
2,2-dimethylhexane
0.20
10
NBP_235
1.84
2-methylheptane
2.22
11
NBP_261
2.10
2,4-dimethylheptane
1.25
12
NBP_286
2.42
2,3-dimethylheptane
2.90
13
NBP_311
2.81
2,6-dimethyloctane
3.25
14
NBP_336
3.12
5-methylnonane
0.00
15
NBP_361
3.14
n-butylcyclohexane
5.17
16
NBP_386
3.14
n-undecane
2.64
17
NBP_411
3.08
n-dodecane
5.20
18
NBP_436
2.83
n-hexylbenzene
0.00
19
NBP_461
2.64
n-tridecane
3.23
20
NBP_486
2.55
n-tetradecane
3.67
21
NBP_511
2.47
n-pentadecane
2.45
22
NBP_536
2.41
n-hexadecane
3.30
23
NBP_561
2.36
n-decylbenzene
0.21
24
NBP_587
2.30
n-heptadecane
1.40
25
NBP_612
2.28
n-octadecane
2.65
26
NBP_637
2.31
n-nonadecane
2.77
27
NBP_662
2.30
n-uneicosane
4.54
28
NBP_687
2.22
n-dodecosane
0.00
29
NBP_712
2.09
n-tricosane
3.51
30
NBP_737
1.91
n-tetracosane
0.00
31
NBP_762
1.73
n-pentacosane
3.06
32
NBP_787
1.62
n-heptacosane
1.19
33
NBP_825
3.03
NBP_821
2.78
34
NBP_875
2.90
NBP_868
3.02
35
NBP_925
2.84
NBP_919
2.90
36
NBP_975
2.80
NBP_970
2.86
37
NBP_1025
2.76
NBP_1021
2.81
38
NBP_1075
2.73
NBP_1072
2.79
39
NBP_1125
2.71
NBP_1123
2.77
40
NBP_1175
2.73
NBP_1175
2.78
41
NBP_1251
5.65
NBP_1251
5.65
42
NBP_1372
8.64
NBP_1372
8.64
Appendix: Detailed Specifications for the Example
All plant data used for the simulation of the oil refining example originated from the HYSYS.Plant documentation (Hyprotech, 1998) and the overview is given by Tables A.1 - A.7.
Table A.1. Stream parameters for crude oil.
Temperature, oC
Pressure, kPa
232.2
517.1
Hot Crude
343.3
448.2
Tower Feed
338.8
448.2
Stream Preheated Crude (flowrate 662.4 m3/h)
Table A.2. Configuration of the atmospheric crude column.
Unit operation
Theoretical Stages
Main tray section
29
Condenser (for the main
1
tray section, partial) Kerosene side stripper
3
Reboiler for the kerosene
1
side stripper Diesel side stripper
3
AGO side stripper
3
Total
40
Table A.3. Connectivity for the Atmospheric crude column.
Stream
Description
From unit operation,
To unit operation, stage
stage TowerFeed
hot crude fed to the
-
main tray section, 28
-
main tray section, 29
column BottomSteam
steam fed to the bottom of the column
Naphtha
naphtha product
condenser
-
WasteH2O
waste water
condenser
-
OffGas
overhead vapor product
TrimDuty
energy stream - trim duty
-
main tray section, 28
Residue
crude atmospheric residue
main tray section, 29
-
Kerosene
kerosene product
reboiler for the
-
-
kerosene SS Diesel
diesel product
diesel SS, 3
-
AGO
atmospheric gas oil
AGO SS, 3
-
-
reboiler for the kerosene
product KeroSS_Energy
kerosene SS reboiler duty
SS KeroSS_Draw
liquid draw stream
main tray section, 9
kerosene SS, 1
KeroSS_Return
vapor return stream
kerosene SS, 1
main tray section, 8
DieselSS_Draw
liquid draw stream
main tray section, 17
diesel SS, 1
diesel SS, 1
main tray section, 16
DieselSS_Return vapor return stream
AGOSS_Draw
liquid draw stream
main tray section, 22
AGO SS, 1
AGOSS_Return
vapor return stream
AGO SS, 1
main tray section, 21
DieselSteam
steam fed to the bottom of
-
diesel SS, 3
-
AGO SS, 3
the diesel SS AGOSteam
steam fed to the bottom of the AGO SS
PA_1_Draw
draw for pump around 1
main tray section, 2
pump around cooler 1
PA_1_Return
return for pump around 1
pump around cooler 1
main tray section, 1
PA_1_Q
cooler 1 duty
pump around cooler 1
-
PA_2_Draw
draw for pump around 2
main tray section, 17
pump around cooler 2
PA_2_Return
return for pump around 2
pump around cooler 2
main tray section, 16
PA_2_Q
cooler 2 duty
pump around cooler 2
-
PA_3_Draw
draw for pump around 3
main tray section, 22
pump around cooler 3
PA_3_Return
return for pump around 3
pump around cooler 3
main tray section, 21
PA_3_Q
cooler 3 duty
pump around cooler 3
-
Table A.4. Parameters of column internal material streams.
Stream
Flowrate, m3/h
Naphtha
132.5
OffGas
0
Kerosene
86.12
Diesel
112.6
AGO
33.12
PA_1_Draw
331.2
PA_2_Draw
198.7
PA_2_Draw
198.7
Table A.5. Parameters of column steam supply.
Parameter
Stream Flowrate, kg/h
Temperature, oC
Pressure, kPa
BottomSteam
3402
190.6
1034.2
DieselSteam
1361
148.9
344.7
AGOSteam
1134
148.9
344.7
Table A.6. Column duty streams.
Stream
Duty, kJ/h
KeroSS_Energy
7.913×106
PA_1_Q
-5.803×107
PA_2_Q
-3.693×107
PA_3_Q
-3.693×107
Table A.7. Various column parameters.
Parameter
Value
Condenser pressure, kPa
135.8
Condenser pressure drop, kPa
62.1
Bottom stage pressure, kPa
225.5
Kerosen SS boil up ratio
0.75
Overflash specification, %
3.5
(or tray net liquid flow from stage 27)
(23.19 m3/h)
New Approach to the Characterization of Petroleum Mixtures Used in the Modelling of Separation Processes Egon Eckert and Tomáš Vaněk
Figures
Tb TI+1
Ti+1 Tbi Ti
T2
T1
0
xi
Φ
1
Figure 1: Definition of a pseudo-component and of its content in the substitute mixture from the TBP curve.
450 400
M [kg/kmol]
350 300 250 200 150 100 50 0
0
100
200
300
400
500
600
700
800
Tb [K] Figure 2. Plot of molecular weights vs. normal boiling points of hydrocarbon components from the HYSYS.Plant database.
350 300
API gravity
250 200 150 100 50 0 -50
0
100
200
300
400
500
600
700
800
Tb [K] Figure 3. Plot of API gravities vs. normal boiling points of hydrocarbon components from the HYSYS.Plant database.
Tb TI+1
Tbi
Ti+1 Ti T2
T1
0
Φ iL Φ i Φ iR
1
Φ
Figure 4. Situation after the selection of a real component in the ith primary temperature interval (Ti,Ti+1).
Tb
Tbi Tbi-1
0
Φ iL−1 Φ i −1 Φ iR−1 Φ iL Φ i Φ iR
1
Φ
Figure 5. Construction of intervals of fraction distilled in the second phase of the algorithm attached to real components selected in the first phase.
Figure 6. Simplified flowsheet for the crude oil processing example.
Figure 7. Subflowsheet for the atmospheric column.
400 350
M [kg/kmol]
300 250 200 150 100 50 300
350
400
450
500
550
600
650
700
Tb [K] Figure 8. Phase portrait for the molecular weight ( ) compared with values retrieved from the database for real components (£).
90
80
API gravity
70
60
50
40
30 300
350
400
450
500
550
600
650
700
Tb [K] Figure 9. Phase portrait for the API gravity ( ) compared with values retrieved from the database for real components (£).
1100 1000 900
Tb [K]
800 700 600 500 400 300 200 0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Φ Figure 10. Measured TBP curve (● ) compared with theoretical TBP curve for the substitute mixture gained by the new approach (£).
900
M [kg/kmol]
750
600
450
300
150
0 0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Φ Figure 11. Measured characterization curve for the molecular weight (● ) compared with the theoretical curve for substitute mixture gained by the new approach (£).
140 120
API gravity
100 80 60 40 20 0 0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Φ Figure 12. Measured characterization curve for the API gravity (● ) compared with the theoretical curve for substitute mixture gained by the new approach (£).
1100 1000 900
Tb [K]
800 700 600 500 400 300 200 0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Φ Figure 13. Comparison of theoretical TBP curves for fractionation products (bottom to top: Offgas, Naphtha, Kerosene, Diesel, AGO, Residue) employing the traditional (¢) and the new (£) approaches to characterization.
Egon Eckert1 and Tomáš Vaněk2
Department of Chemical Engineering, Institute of Chemical Technology, Prague, Technická 5, 166 28 Prague 6, Czech Republic
Characterisation of complex mixtures is a common tool especially in oil processing industry. Characterisation procedures result in experimentally gained characterisation curves, but for the simulation of industrial processes the definition of a substitute mixture is required. Traditionally, a system of pseudocomponents is derived from the TBP (True Boiling Point) characterisation curve, but there are a number of disadvantages, e.g. the physical properties of pseudocomponents must be estimated by unreliable empirical methods. The new approach to the characterisation of complex mixtures is based on representing the original mixture by a system of real components. Such substitute mixture is fully defined, it has a chemical character, and physical properties can be simply retrieved from databases. Utilisation of a substitute mixture of real components in the simulation of crude oil processing proved that the new approach could replace the traditional one in normal boiling temperature ranges where real components are available. Both approaches could be also easily combined.
Keywords: Refinery processes, Oil processing, Characterisation procedures, Complex mixtures, Pseudocomponents
1
[email protected]; 2 [email protected]
1.
Introduction Mixtures containing an extremely large number of various components can be often
encountered in industrial chemical technologies, particularly in oil processing and refining. Briesen and Marquardt (2003, 2004a, 2004b) analysed thoroughly the past, present and future trends in this branch of industry and pointed out the need to increase the accuracy in the modelling of oil refining processes. There are direct economical and environmental consequences of any progress in this point and improved treatment of complex mixtures is probably the most promising direction. In order to deal with such mixtures in modelling and simulation calculations, it is necessary to simplify the problem by utilising a substitute mixture possessing reasonably lower number of components or to use an alternative representation of the mixture. For this purpose the continuous thermodynamics can be employed (see e.g. Rätzsch & Kehlen, 1983), or, more recently, the wavelet-Galerkin discretization has been proposed (Briesen & Marquardt, 2003, 2004a, 2004b). While continuous thermodynamics has received only a little attention in industrial practice, the adaptation of wavelet-Galerkin discretization method for the modelling of unit operations, where complex mixtures are processed, is intensively being studied. Its main advantage is the possibility to tune the representation of the mixture by means of adaptive control of the problem discretization. On the other hand, this approach uses a non-trivial mathematical background and also the standard models of unit operations must be reformulated in order to incorporate the continuous representations (distributions) of some model variables, e.g. of the vector of composition (Briesen & Marquardt, 2003, 2004a, 2004b). Moreover, physical properties dependent on the composition (e.g. K-values) must be also converted to distribution-based functions. Such methods can be employed only for physical processes as distillation or absorption, but not for processes with chemical reactions. Since the last decade of 20th century, methods for the "reconstruction" of the chemical composition have been developed presuming that some other information was available
beside the usual distillation curve. It might be data from elemental analysis, gas chromatography, mass spectrometry, 1H and 13C NMR analyses, etc., (e.g. Whitson & Brulé, 2000). When using substitute mixtures in steady-state or dynamic simulation, the classical unit operation models can be employed. In fact, in simulation programs the solution engine deals equally with real components as well as with pseudocomponents if properly defined during the data input phase. The number of components used for the substitute mixture is usually up to 102, which is considerably lower than for original complex mixtures, i.e. 104 - 107 or more in case of crude oils. It might be felt that in the age of powerful computing machinery the number of components in chemical engineering calculations could be practically unlimited. Nevertheless, there are at least two good reasons why to keep lower number of components. First, the dimension of unit operation models is dependent on the number of components involved and especially equation-oriented simulators could still run into troubles according to memory requirements and internal limits. Second, it is very hard to analyse the results of simulation run when the number of components is high, e.g. thousands or more, and no new information could be obtained. Probably, there would be subsets of components behaving almost equally, the content of each being on the ppm level. It could be noticed that also for well-defined mixtures in practical calculations it is often desired to decrease the number of components or even pseudocomponents. The method called lumping is based on representing groups of components with close boiling points and/or some other properties by a selected single member component inheriting in the mixture the "weight" of the entire group (Riazi, 2005). Montel & Gouel (1984) suggested to optimise the lumping scheme for the substitution of a group of known components in order to preserve the PVT behaviour. The approach based on pseudocomponents had been developed quite a long time ago (Edmister, 1955; Katz & Brown, 1933) and first was used for flash calculations on an early computer (Hariu & Sage, 1965). As the main advantage it was appreciated that the characterisation procedure was non-iterative. It is still widely accepted as a convenient method in the simulation of separation equipment, but a number of problems arise. Above all:
•
The chemical character of components forming the mixture could also play a role in chemical reactions occurring in the studied processes. For pseudocomponents we can't define any chemical character.
•
A pseudocomponent is primarily defined only by its (pseudo-) boiling point and by some additional parameters, mostly by specific gravity, molecular weight or viscosity. All other physical properties, e.g. needed for simulation calculations, must be estimated. Unfortunately, the reliability of common estimation methods (listed for example by Whitson & Brulé, 2000) for critical properties, the acentric factor, etc. is rather low according to the results of testing published for example by Twu (1984), Lindqvist et al. (1994), Riazi (2005).
•
For pseudocomponents we cannot use group contribution methods (e.g. UNIFAC) requiring information about the molecular structure of compounds in order to estimate some parameters (e.g. binary interaction parameters for vapour - liquid equilibrium).
•
The information about the type of the mixture, e.g. if paraffinic or aromatic compounds are prevailing, or about the type of some of its important components (e.g. polar compounds) couldn't be easily utilised.
•
Arbitrary combinations of pseudocomponents and compounds identified in the original mixture are not supported in commercial simulation programs. That is, it is not possible to place a real component into the middle of the temperature range used for the definition of pseudocomponents without knowing its content.
When modelling distillation and absorption processes for complex hydrocarbon mixtures within simulation programs, the original mixture is traditionally substituted by a mixture comprising two different groups of components. For the light end, real components usually up to C5 are employed if they are present and if their content can be determined. For the remaining part, a system of pseudocomponents is derived from certain global characterisations of the mixture based on its distillation behaviour. This approach is implemented in recent major commercial simulation
programs as ASPEN Plus, HYSYS and PRO/II (ASPEN Plus and HYSYS are registered trademarks of Aspen Technology and PRO/II is a trademark of SimSci-Esscor). Beside these standard features, there are most recent attempts to implement other approaches to oil and gas processing, e.g. available with HYSYS 3.2 onwards as HYSYS Upstream™ Option. The usage of pseudocomponents has after all one important advantage: when defined and equipped by a set of estimated physical (pseudo)properties, pseudocomponents could be in simulation programs treated as any other real components (except in processes with chemical reaction). This is also the main presumption for the alternative way how to establish a feasible substitute mixture as suggested Ba et al. (2003). It has been shown that the substitute mixture composed of suitable real components can be used in simulation calculations instead of an alternative mixture of pseudocomponents with comparable results (Eckert & Vaněk, 2003). The primary intention was to overcome the main disadvantages of pseudocomponents mentioned above, namely the need to estimate their properties. It is possible to retrieve property data from the database, which is used as the source of real components. At the same time, wide variety of thermodynamic models and packages of compatible methods, typically within simulation programs, can be used. There is no need to limit to equation-of-state (EOS) models but other types of models, especially for the description of the phase equilibrium, can be employed, even those requiring some interaction parameters. Moreover, the mixture of real components exhibits a chemical character, which can be utilized when modelling complex reaction schemes. For example, Bělohlav et al. (2005) recently used substitute mixtures of real components for the prediction of yields from pyrolysis reactors. Despite the simplicity of the new approach, it has not been used before, e.g. no attempt to employ real components for this purpose is mentioned even in the most current Riazi's monograph (Riazi, 2005). The principles of the new approach makes it directly usable in simulation programs with current content of their databases of physical properties and for current models of unit operations.
Further in this text, we shall recapitulate the two approaches employing substitute mixtures in order to explain the difference between the appropriate algorithms. The verification of the new approach is then provided on a non-trivial simulation case - crude oil refining.
2.
Characterisation of mixtures using pseudocomponents Normally, two steps are needed to define a system of pseudocomponents for a complex
mixture. First, some standard characterisation curve must be experimentally obtained, preferably the TBP (True Boiling Points) curve, which is directly used in the next step. The TBP curve represents the dependence of temperature measured in the head of a laboratory batch column on mass or volume fraction distilled. The column should have a high number of stages and perform at large reflux ratio (API, 1992). There are other possibilities, e.g. the TBP curve can be substituted by its more accurate chromatographic equivalent SIMDIST or by transformation of curves resulting from some other characterisation procedures (ASTM D86, EFV - Equilibrium Flash Vapour) using empirical correlations (API, 1992). The later case is not too reliable and can lead to a considerable deviation of the calculated TBP curve from the experimental curve (Ba et al., 2003). In the second step the range of boiling points of the TBP curve is cut in order to obtain non-overlapping temperature intervals (Ti, Ti+1), for i=1,...,I, continuously covering the entire temperature range. There are various possibilities how to choose these intervals - some recommendations were published, for example, in Whitson & Brulé (2000). Usually, it is sufficient to use about 15 K for normal boiling points up to 700 K, about 30 K within 700 and 950 K and about 50 K for higher boiling mixtures as it is done in the HYSYS simulation program when using automatic cutting option (see Hyprotech, 1998). Each temperature interval represents one pseudocomponent with normal boiling point given by the arithmetic or, more precisely, the integral mean temperature over the corresponding interval of fraction distilled - see Figure 1. The relevant definitions of arithmetic and integral mean temperatures are given by Eqns (1) and (2) respectively.
Tbi =
Tb (Φ iR ) + Tb (Φ iL ) , 2
1 Tbi = R Φ i − Φ iL
ò
ΦiR
ΦiL
i = 1,..., I
Tb (Φ ) dΦ ,
i = 1, ..., I
(1)
(2)
As also illustrated by Figure 1, the interval of fraction distilled (Φ iR − Φ iL ), i = 1, ..., I determines at
the same time the relative contribution of each pseudocomponent to the mixture, which is used to define its composition. For the consequent usage of the substitute mixture of pseudocomponents it is necessary to supply the same structure of physical properties as for real components. Usually, it is not difficult to measure besides the TBP curve also the density and/or viscosity of individual fractions collected in the laboratory column during the distillation test. Moreover, for each fraction the mean molecular weight can be estimated. Therefore, if we use pseudocomponents to represent these fractions then two to four data items are available for each pseudocomponent: the mean normal boiling point and one to three other properties, which can be used to trigger a series of empirical estimation procedures delivering the remaining set of properties - critical properties, acentric factor etc. (see e.g. Riazi, 2005; Whitson & Brulé, 2000). Unfortunately, we should be aware of low reliability of such methods as it was mentioned in the introductory part of this contribution. It is relatively simple to measure global parameters of a complex mixture, i.e. bulk properties as the density, molecular weight, refraction index or viscosity, and this is often delivered with other measurements. Unfortunately, their direct use for the derivation of a substitute or model mixture is known only for the density where it is possible to construct under certain conditions an approximation of the characterisation curve from the Watson factor (Wauquier, 1995). When the global density of the mixture is known and the Watson factor KW =
( 1.8Tb )1 / 3 S
(3)
can be assumed to be constant then it is possible to get an approximation of the density characterisation curve (Wauquier, 1995). This could be an alternative to density data measured for each fraction during the characterisation experiment.
3.
Characterisation of mixtures using real components
The main disadvantages of the approach based on pseudocomponents can be overcome if we employ real components to form the substitute mixture. Of course, the selection of suitable real components and the derivation of the substitute mixture must follow certain criteria and an appropriate algorithm must be defined. We shall use two consecutive phases, i.e. a set of components to be present in the substitute mixture must be first selected and then the composition of their mixture will be adjusted. The prerequisites for the algorithm are as follows:
a. The TBP curve measured for the original complex mixture is available Tb = Tb (Φ )
(4)
where Tb is the measured "True Boiling Point" temperature and Φ is the mass or volume fraction distilled. b. Some other characterisation curve is available, e.g. all or some of dependencies
M = M (Φ )
(5)
ρ = ρ (Φ )
(6)
η = η (Φ )
(7)
where M is the molecular weight, ρ the liquid density (eventually specific or API gravity) and
η is the liquid viscosity. The liquid density curve can be also obtained approximately using the Watson factor as mentioned above. Instead of using Eqns (5)-(7) directly, it is better to convert them into "phase portraits" by eliminating the mass or volume fraction distilled and establishing direct relation between these properties and the boiling point temperature:
M = M (Tb )
(8)
ρ = ρ (Tb )
(9)
η = η (Tb )
(10)
c. The overall temperature range is divided (cut) into a system of non-overlapping temperature intervals continuously covering the range. There are several possible approaches similarly to the definition of pseudocomponents - an equidistant grid, adaptive grid taking into account the shape of the curve or a grid based on boiling points of homological series of alkane compounds - see, e.g., Table 5.2 in Whitson & Brulé (2000). d. A sufficiently large database of chemical components and their physical properties is available. Detailed requirements will be discussed further.
3.1
Selection of components The basic assumption is that a suitable database is available. The "quality" of such database can
be expressed not only by the reliability of data included but also by the extent of the database. Here we take primarily into account the number of components but on its own this is not sufficient. It is important to have satisfactory representation for important families of chemical compounds, in our case especially hydrocarbons or detailed groups as e.g. parrafines or aromatics. At the same time, the range of normal boiling points for these compounds must be wide and the normal boiling points should cover the range uniformly and densely. This is, of course, an ideal situation, which is more or less fulfilled by databases from various sources. In "open" literature we can find, for example, the API database (API, 1992) with 478 hydrocarbons up to C30, normal boiling points up to 720 K and molecular weights up to 420 kg/kmol. Generally, normal boiling points of higher hydrocarbons are very difficult to obtain as the thermal decomposition takes place (Kopsch, 1995). Accordingly, data for higher hydrocarbons (C18+) in databases are frequently estimated or their origin is ambiguous. Nevertheless, their long-term usage in many chemical engineering calculations
giving realistic results gives certain guaranty that also the substitute mixture of real components selected from such databases can sufficiently model the original complex mixture. The internal database of the HYSYS simulation program (Hyprotech, 1998), which we used for test calculations, incorporates 521 hydrocarbons and covers approximately the same range of normal boiling points and molecular weights as the API database. The source of data is not explicitly specified and probably the same limitations can be expected as for API data. Thermal decomposition is also one of the reasons, why the number of available hydrocarbon compounds rapidly decreases with increasing normal boiling point as demonstrated by Figures 2 and 3. This situation leads to a poor chance to find a suitable real component to represent higher boiling point temperature ranges. Unfortunately, the mixtures in oil processing contain a significant portion of higher boiling compounds forming the so-called "heavy end". Some components can exhibit normal boiling point higher than 1000 K and molecular weight exceeding 800 kg/kmol. Normal boiling points and densities, as the minimum information needed, can be found, e.g., in Beilstien database (MDL Information systems, 2005), for about 1850 hydrocarbons up to C42, where only two components are available - 1-cyclohexyl-hexatriacontane a 1-phenylhexatriacontane. Figures 2 and 3 show another interesting feature. With increasing temperatures the range of molecular weights of components having approximately the same boiling point is getting wider, see for example the region around 700 K in Figure 3. There are a number of other databases, which could be potentially used for the same purpose, e.g. DIPPR (Design Institute for Physical Properties at Brigham Young University, 2000) or the suite of internal databases attached to the ASPEN Plus. The reasons why we did not employ these databases are technical as the program providing the selection of components needs a self-standing "flat" data file. Our approach, initially designed for exclusive usage of real components in the substitute mixture, can be extended to combine real components for lower and moderate boiling point temperatures with pseudocomponents for the heavy end. Nevertheless, this demonstrates the potential of the new approach.
The procedure for the selection of real components for the substitute mixture can be made flexible in order to exploit all information about the original mixture. As stated above, if the chromatographic analysis of the mixture is available it could be also a basis for the formation of a substitute mixture. Certain components can be clearly identified and directly added to the list of components in the substitute mixture. If no chromatographic analysis is available but the TBP curve is known, all components for the substitute mixtures can be selected according to the following algorithm:
Step 1: For each primary temperature interval (prerequisite c.) a set of candidate components is
selected from the database. The criterion is that each component selected must have its normal boiling point within the considered temperature interval. If needed, filtration conditions can be set, e.g. reflecting the requirements to include only some families of components or, on contrary, to exclude another components or families. For example, it is known that no olefin components are present in crude oils (Petrov, 1987). On the other hand, at least one component should be available for each interval. If not, then either a set of wider intervals must be used or the interval can be represented by a pseudocomponent. The combination of real components and pseudocomponents is necessary for higher boiling mixtures. Step 2: Exactly one component is selected for each primary temperature interval from the set
of candidate components by comparing their physical properties. We know the normal boiling point of each candidate component and from phase portraits (8)-(10) the "desired" values can be derived and compared with values retrieved from the database. The simplest way how to combine the deviations for different properties is to use a weighted sum of relative differences. The criterion for the selection is then defined by K
å k =1
wk
ζ r ,k ,c − ζ m ,k ,c ζ m ,k ,c
→ min c
(11)
where K is the number of measured properties and ζ m ,k ,c , ζ r ,k ,c are the measured and from database obtained (calculated or simply retrieved) values of property ζ respectively. The expression is calculated for each candidate component, c = 1,...,Ci and the component with the lowest value of criterion (11) is chosen to represent the interval. There are degrees of freedom in the choice of weight factors wk , which can reflect the precision of measurements or some other demands. It should be noted that measured viscosity and density in fact represent properties of discrete mixtures (fractions), which can be hardly derived from properties of contained pure components according to non-trivial and unreliable mixing rules. Sometimes no other curves than the TBP curve are available. Step 2 of the algorithm can be then replaced by choosing a component with its normal boiling point being closest to the mean temperature of the appropriate temperature interval. This is an emergency solution but it can be expected that the selected real component would be close to a pseudocomponent defined traditionally while all properties of the real component are directly available from the database. The result of this phase of the algorithm is simply a list of real components together with their normal boiling points and other properties. If it is desired for some reason to include other components into the substitute mixture, it can be done now. Such an obvious reason is, for example, the confirmed presence of a particular component or component type in the original mixture. Especially polar compounds are in the focus since they strongly affect the phase equilibrium in multiphase systems. A compound can be added to the substitute mixture with or without information about its amount in the original mixture. Both possibilities are aided in the second phase of the algorithm.
3.2
Determination of the composition of a substitute mixture The problem is that the normal boiling point of each selected component can fall anywhere in
the primarily defined temperature interval (Ti, Ti+1), see Figure 4, and no longer it is a mean temperature of this interval. If a model of the experimental characterisation procedure were
available then we could abandon the principle of mean temperatures and this step would consist in a repeated evaluation of the characterisation curve from the model for varying composition of the mixture until a satisfactory match with the experimental characterisation curve is reached. This kind of an optimisation can be easily employed, for example, for the EFV curve (Eckert, 1999). When the TBP curve is available then the procedure suggested by Ba et al. (2003) can be used. Figure 5 illustrates the principles of the procedure, which at first glance could be found similar to the method used to derive the composition of a mixture of pseudocomponents. The presumption that each component should represent an interval of fraction distilled, where its normal boiling point is the mean temperature according to Eqns (1) or (2), is preserved. The difference is that we cannot expect that the intervals could be generally chosen to cover the whole range without gaps and/or overlapping. This is automatically fulfilled for pseudocomponents where the procedure starts from the other end - the mean temperatures are calculated for temperature intervals initially chosen to continuously cover the entire temperature range without overlapping. Nevertheless, in the case of real components the idea is to distribute the intervals to make them cover the range as much as possible with minimum gaps and overlapping. This is an optimisation task where the objective function can be defined as follows:
( ) å (Φ
F ΦL =
I +1
i = LE +1
R i −1
− Φ iL
)
2
→ min
(12)
For each interval the starting and end fractions distilled are denoted by the superscript L (= Left) and R (= Right). Generally, we can utilise the information about light-end components, which can be usually identified in the original mixture together with their composition, and incorporate them into the substitute mixture. The index of the last light-end component is therefore denoted by LE R and the corresponding fraction distilled by Φ LE . On the other end of the range, I +1 can be denoted
by HE (heavy-end), even if no heavy-end components were present and by definition L L L Φ HE = Φ IL+1 = 1 . The reason why only the elements of the vector Φ L = {Φ LE +1 ,...,Φ I } are considered
as independent optimisation variables in (12) is that Eqns (1) and (2) can be treated as implicit
( )
equations for Φ i , i = LE + 1,.., I , i.e. Φi = Φi Φi . Alternatively, the reversed relation R
R
R
L
( )
Φ i L = Φ i L Φ i R could have been used with Φ i R as the set of optimisation variables. It is reasonable to define bounds for the values of Φ i in order to avoid getting some unreal solutions and the L
following set of constraints was considered:
Φ iL ≤ Φ iL+1 , i = LE + 1,..., I
(13)
Φ i−1 ≤ Φ ≤ Φ i+1 , i = LE + 1,..., I L i
Optimisation problem (12) together with constraints (13) can be solved using some standard package, e.g. MINOS (Murtagh & Saunders, 1995) as in our case. In order to obtain a consistent composition of the substitute mixture it is necessary to convert
(
the intervals Φ i ,Φ i L
R
) resulting from the optimisation to mass or volume fractions. In ideal case,
the intervals would exactly cover the entire range, the value of the object function (12) would be zero and the following condition would be fulfilled:
å (Φ I
i = LE +1
R i
)
L R − ΦiL = Φ HE −Φ LE
(14)
Then the length of each interval can be directly taken as the appropriate mass or volume fraction. Nevertheless, this is not usually true according to possible overlapping of the intervals and existence of areas not covered with any interval. The simplest way how to get a consistent
(
composition derived from the lengths of intervals Φ i ,Φ i L
R
) is to "normalise" the vector of mass or
volume fractions:
(
x j = Φ −Φ R j
L j
)(
L Φ HE
) å (Φ I
R − Φ LE
R i
)
− Φ iL ,
j = LE + 1,..., I
(15)
i = LE +1
A similar procedure can be used when some of the real components were identified in the original mixture and their relative amounts, i.e. mass or volume fractions, are known. The second phase of the algorithm was implemented as self-standing program incorporating the MINOS (Murtagh & Saunders, 1995) optimisation package.
4.
Application of the method
In order to approve the suggested approach it was reasonable to simulate real industrial processes employing substitute mixtures. Unfortunately, the data measured, for example, on distillation columns and published in open literature are almost always for binary mixtures and small-scale columns. Data from major distillation columns in crude oil processing are very rarely available. Therefore, an alternative way is to compare at least the two different approaches to substitute mixtures on the same simulation problem, once using a substitute mixture composed of pseudocomponents and then a mixture of real components. An important point is to define some methodology how to compare different substitute mixtures. In a former contribution (Eckert & Vaněk, 2003) we used the concept of a "theoretical TBP curve" as curve of a staircase shape reflecting the distribution of normal boiling points of individual components from the mixture. Such curve we could get in a batch distillation column with the number of stages approaching infinity and very high reflux ratio. In its discrete version the j =i −1
theoretical TBP curve is composed of points with co-ordinates (
åx
j
+ xi / 2 ; Tbi) for i = 1,...,I,
j =1
taking into account for simplicity the arithmetic mean values defined by Eqn. (1). Analogical theoretical curves can be constructed for molecular weight and density, eventually API gravity or some other properties. Theoretical TBP curves were constructed for examples of separations in a distillation and absorption columns (Eckert & Vaněk, 2003) and the match between results obtained by both approaches was excellent, thus showing that substitute mixtures of real components can replace mixtures of pseudocomponents with all the extra benefits, which this approach brings. For these examples only the measured TBP curves were available as input information. The example presented further is based on the "Refining Tutorial" case delivered with the simulation program HYSYS.Plant version 2.1 (abrev. HYSYS) and at the same time this program
has been used for all calculations. The advantage is that HYSYS incorporates a relatively extensive database of hydrocarbons and it also enables to define pseudocomponents in a built-in tool called "Oil Environment".
Example (Hyprotech, 1998) Crude oil is processed in a fractionation facility to produce naphtha, kerosene, diesel, atmospheric gas oil and atmospheric residue products. The crude oil is characterised by laboratory assay data in Tables 1.-3. The TBP curve is accompanied by the dependence of molecular weight on the liquid volume fraction distilled and a separate assay on API gravity is also available. A light end is considered and its composition is known. The main flowsheet for the process in consideration is shown in Figure 6. Preheated crude is initially fed to a pre-flash drum to separate vapours from the liquids, which are further heated in a furnace. The pre-flash vapours are re-combined again with the hot crude from the furnace and the resulting stream is then fed to the atmospheric crude column for fractionation. Detailed description and specifications are presented in the Appendix. The icon of the distillation column in Figure 6 represents in fact a complex separation process, which can be in the HYSYS program further represented by an embedded flowsheet, i.e. subflowsheet in HYSYS terminology. The appropriate subflowsheet for our example is shown in Figure 7. The PengRobinson property package was used in HYSYS for the calculation of physical properties. In the "Oil Environment" the standard setting of estimation methods was used, i.e. Twu critical property correlation for molecular weight (Twu, 1984), constant Watson factor method for specific gravity and Lee-Kesler method for critical properties, the acentric factor and ideal enthalpy (Kesler & Lee, 1976). It was intended to solve the simulation case using both methods for the characterisation of the process feed (Preheat Crude stream). Particularly, Table 4 globally summarises how the substitute mixtures were assembled. Temperature ranges defined in the first column of the table were for pseudocomponents divided uniformly to intervals of the same width. For real components this was
also done but, as we know from the description of the algorithm above, it serves only for the initial selection of subsets of real components as candidates for representing these intervals. For both substitute mixtures the same light-end (Table 1) was supposed. Unfortunately, higher temperature ranges had to be covered by 10 pseudocomponents, as there were only few hydrocarbons in the HYSYS database with boiling points exceeding 426.7 oC. Our program used for the selection of real components, accepts input data in various forms and physical units and provides necessary conversions. For our example the conversions of characterisation curves from tables 2 and 3 to phase portraits expressed by Eqns (8) and (9) were done using a piece-wise linear interpolation. The phase portrait for the API gravity has been extended to include the light end components. The reason is that this curve bends at the boundary between light end and the remaining part of the mixture, which affects interpolation or extrapolation operations needed in the algorithm. Then for each temperature interval a set of candidate components was chosen from the HYSYS database having normal boiling points within this interval. This primary selection was limited to hydrocarbon compounds only but also compounds known to be missing in oil fractions (e.g. olefins, cycloolefins) were excluded. Technically, this is enabled by registering the chemical family of compounds and setting an "oil" flag on or off for each component in the database. The most suitable component for each interval was found according to criterion (11) for K = 2, i.e. taking into account the molecular weight and API gravity. The weight factors given to the contributions of both properties were simply set to be one. The resulting list of real components for the substitute mixture is presented in Table 5 and the comparison of the measured and retrieved values of both considered properties are depicted in Figures 8 and 9 as phase portraits. Table 5 contains also the information about the number of candidate components for each primary temperature interval. It is apparent that their number rapidly decreases when approaching higher temperatures. It is also interesting that starting from interval number 16 almost all components selected belong to a monotone series of alkanes. Only components number 18 (n-hexylbenzene) and 23 (n-decylbenzene) do not follow this fact and show
larger deviation from the phase portrait as documented by Figures 8 and 9. Another component showing larger deviation in Figure 9 is n-butylcyclohexane in interval 15. In order to improve the selection it could be possible, for example, to try larger temperature intervals or their different distribution, but it would disable to compare the new approach with the traditional one. Therefore, the actual result of selection was left unchanged and passed to the second phase of the algorithm. After selecting a unique component for each primary temperature interval the actual intervals
(Φ
L i
,Φ i
R
) attached to each component were computed in the second phase of the algorithm and the
composition was derived using Eqn. (15). Table 6 contains the complete overview of resulting substitute mixtures including the light and heavy ends. Pseudocomponents were generated in HYSYS Oil Environment and their names chosen by HYSYS reflect the mean temperatures of appropriate intervals (in deg F). For the new approach the interpretation according to HYSYS could be that the four light-end components (originally 1.13 vol. %) together with the system of selected real components would form an extended light-end (in our example it will comprise 32 components and 63 vol. % of the mixture). The measured and theoretical TBP curves of the Preheat Crude are compared in Figure 10. The comparison of experimental additional curves with properties of real components selected for substitute mixture is depicted in Figures 11 and 12. In these Figures it can be also observed that in the region of higher boiling temperatures program HYSYS is not so successful in the choice of pseudocomponents and a considerable deviation between the experimental and estimated values of the molecular weight as well as API gravity is apparent. The simulation calculations of the entire crude oil fractionation process employing both approaches to the characterisation of complex mixtures run without any problems. The excellent match between both approaches can be demonstrated by Figure 13, where the comparison is provided for all the products of fractionation, which have distinct mean boiling points. The plotted theoretical TBP curves were recalculated in order to eliminate water. In fact, the TBP curves reflect the match in the composition of product streams, as the most important parameter, after processing of a single feed stream in a number of unit operations within a relatively complex technology with
several recycle streams. Also for other parameters of the product streams, as flowrates, temperatures and enthalpies, the results of simulation proved very nice match. It can be noted that these results were reached despite the fact that the database used is relatively small compared to other data sources. Moreover, the content of some of the real components selected into the substitute mixture resulting from the second phase of the algorithm is practically negligible (see Table 6 - components number 6, 14, 18, 28, 30).
5.
Conclusions
The results presented in this paper allow us to say that the new simple approach to the characterisation of complex mixtures is fully acceptable even for simulation calculations of largescale and complex processes including various mass and heat transfer operations. It can replace the traditional approach based on the definition of pseudocomponents for low and moderate normal boiling points. Of course, there is an important assumption about the availability and sufficient number of real components with normal boiling points in the considered temperature range in the database. If necessary, we can always use a combined approach, as in the example above, adding pseudocomponents for higher boiling temperature range where no real components are available in the database. We can remark that the number of real components not only in the HYSYS but also in other databases we have been dealing with is still extremely low compared to the possible number of all hydrocarbons. It is an interesting fact that considering hydrocarbon molecules with 25 carbon atoms, the number of possible configurations already exceeds 600 millions (Krambeck, 1991) and crude oil may contain a huge number of distinct molecular species in the order 106 (Altgelt & Boduszynski, 1994). Nevertheless, substitute mixtures allow to deal with complex mixtures in modelling and simulation of technological processes without need to dispose with data for such extent of existing compounds. Systematic addition of missing data and inclusion of new components into databases is continuously provided by the vendors of the main databases of physical properties (e.g. Beilstein) and certainly it brings better possibilities for the selection of real
components into a substitute mixture, but even current pallet of available compounds gives good results. This was proved not only by the example presented above but also by some other (Ba et al, 2003; Bělohlav et al., 2005; Eckert & Vaněk, 2003, 2005). There are many benefits when using real components for substitute mixtures instead of pseudocomponents: •
The usage of substitute mixtures with real components can be extended from usual calculations of separation processes to processes with chemical reactions (Bělohlav et al., 2005) as the substitute mixture receives a chemical nature, even if only a "substitute" one.
•
Empirical estimation methods for physical properties are generally not needed. The values retrieved from database are more reliable and precise despite the fact that for higher boiling compounds (approximately C20 and higher) they are predicted only. On the other hand, the knowledge of the molecular structure allows us to use group contribution methods to estimate various physical properties, e.g. the phase equilibrium behaviour of the substitute mixture (and therefore of the original mixture as well).
•
The selection of components can be affected by partial information about components positively occurring in the mixture or about the overall character of the mixture. Particularly, we can recognise in some mixtures components containing nitrogen or sulphur, but there is certain problem with availability of data. For example, the HYSYS database contains about 50 sulphuric compounds the presence of which in oil could be expected. On the other hand, the inclusion of such components into the resulting substitute mixture is very important when modelling separation processes having strong impact on the environment (Eckert & Vaněk, 2005).
The new approach could be also very convenient for the implementation into standard commercial simulation programs where the combination with current proprietary databases and libraries of numerical methods, especially optimisation procedures for the second phase of the
algorithm, could be very profitable. There are also certain possible improvements in the construction of the substitute mixture, e,g. the selection of components an/or the composition can be optimised in order to get best match between experimental characterisation data and results of the modelling of the appropriate characterisation procedure.
Acknowledgement
Authors appreciate the support of the fund MSM 6046137306.
List of symbols
C
number of candidate components in the primary temperature interval
F
objective function
I
total number of real components
K
number of measured properties
KW
Watson factor
M
molecular weight
S
standard specific gravity
T
temperature
x
volume or mass fraction
w
weight in criterion (11)
η
viscosity
Φ
volume or mass fraction distilled
ρ
density
ζ
symbol for property
Subscripts b
at boiling point
c
index of a candidate component
HE
index of the first heavy-end component
i
index of a component or temperature interval
j
index of a component
k
index of a property in criterion (11)
LE
index of the last light-end property
m
measured value
r
value calculated or retrieved from the database
Superscripts L
left edge of an interval
R
right edge of an interval mean value
Literature Cited
Altgelt K., & Boduszynski M. (1994). Composition and Analysis of Heavy Petroleum Fractions. New York: Dekker. American Petroleum Institute (API) (1992). Technical Data Book - Petroleum Refining. 5th ed. Washington: API. Ba A., Eckert E., & Vaněk T. (2003). Procedures for the selection of real components to characterize petroleum mixtures. Chem. Pap., 57, 53. (full text available at http://www.vscht.cz/uchi/procesy/) Bělohlav Z., Zámostný P., Herink T., Eckert E., & Vaněk T. (2005). A Novel Approach for the Prediction of Hydrocarbon Thermal Cracking Products Yields from the Substitute Feedstock Composition. Accepted for publication in Chem. Eng. Technol. (full text available at http://www.vscht.cz/uchi/procesy/)
Briesen H., & Marquardt W. (2003). An adaptive multigrid method for steady-state simulation of petroleum mixture separation processes. Ind. Eng. Chem. Res., 42, 2334. Briesen H., & Marquardt W. (2004a). New approach to refinery process simulation with adaptive composition representation. AIChE J., 50, 633. Briesen H., & Marquardt W. (2004b). Adaptive multigrid solution strategy for the dynamic simulation of petroleum mixture processes. Comp. & Chem. Engng., 50, 633. Design Institute for Physical Properties (2000). DIPPRâ 801. Brigham Young University, U.S.A. Available at http://dippr.byu.edu/product.asp. Eckert E. (1999). Non-traditional Characterization of Petroleum Mixtures in Terms of Selected Components. Collect. Czech. Chem. Commun., 64, 571. (full text available at http://www.vscht.cz/uchi/procesy/) Eckert E., & Vaněk T. (2003). Simulation of separation columns using substitute mixtures. Proc. of the 30th Int.Conf. of SSCHE on CD-ROM. Tatranské Matliare, Slovakia, May 26-30, 2003. (full text available at http://www.vscht.cz/uchi/procesy/) Eckert E., & Vaněk T. (2005). Extended utilisation of the characterisation of petroleum mixtures based on real components. Proc. of the 32th Int.Conf. of SSCHE on CDROM. Tatranské Matliare, Slovakia, May 23-27, 2005. (full text available at http://www.vscht.cz/uchi/procesy/) Edmister W. (1955). Improved integral technique for petroleum distillation calculations. Ind. Eng. Chem., 47(9), 1685. Hariu, O. H., & Sage, R. C. (1969). Crude Split Figured by Computer. Hydrocarbon Processing, 48(4), 143. Hyprotech Ltd. (1998). HYSYS.Plant 2.1 Documentation. Calgary: Hyprotech.
Katz D., & Brown G. (1933). Vapor pressure and vaporization of petroleum fractions. Ind. Eng. Chem., 25(12), 1373. Kesler M.G. & Lee B.I. (1976). Improved prediction of enthalpy of fractions. Hydrocarbon Processing, 55, 153. Kopsch H. (1995). Thermal Methods in Petroleum Analysis. Weinheim: VCH. Krambeck F. (1991). Continuous mixtures in fluid catalytic cracking and extensions, in Sapre A. and Krambeck F. (Eds.), Chemical reactions in complex mixtures. Pages 42-59. New York: Van Nostrand Reinhold. Lindqvist P., Markkanen V., & Happonen V.M. (1994). Simulation of a heavy residue vacuum column. ASPENWORLD '94. November 6-9. Boston, Massachusetts, 1994. MDL Information Systems (2005). Crossfire Beilstein Database. San Leandro, CA: Elsevier MDL. Available at http://www.mdl.com/products/knowledge/crossfire_beilstein/. Montel, F., & Gouel P.L. (1984). A new Lumping Scheme of Analytical Data for Compositional Studies. Presented at the 59th Annual Technical Conference and Exhibition, paper SPE 13119. Houston, Sept. 16-19,1984. Murtagh, B.A., & Saunders, M.A. (1995). MINOS 5.4 User's Guide. TR SOL 83-20R. Department of Operations Research, Stanford University. Stanford CA. Petrov A.A. (1987). Petroleum Hydrocarbons. Berlin, Heidelberg: Springer-Verlag. Rätzsch M., & Kehlen H. (1983). Continuous thermodynamics of complex mixtures. Fluid Phase Equilibria,14, 225. Riazi, M. R. (2005): Characterization and Properties of Petroleum Fractions. Barr Harbor: ASTM. Twu, C.H. (1984). An Internally Consistent Correlation for Predicting the Critical Properties and Molecular Weights of Petroleum and Coal-Tar Liquids. Fluid Phase Equilibria 16(2), 137.
Wauquier, J.-P. (ed.) (1995). Petroleum Refining, Vol. 1: Crude Oil. Petroleum Products. Process Flowsheets. Paris: Éditions Technip. Whitson, C. H., & Brulé, M. R. (2000). Phase Behavior. SPE Monograph Series. Richardson: Society of Petroleum Engineers, Inc.
Tables
Table 1. Light end components and their liquid volume %.
No.
Component
Liq. vol. %
1
i-butane
0.19
2
n-butane
0.11
3
i-pentane
0.37
4
n-pentane
0.46
Table 2. TBP Distillation Assay.
o
Molecular weight,
Liq. vol. % Temperature, C kg/kmol 0
26.7
68
10
123.9
119
20
176.1
150
30
221.1
182
40
275.0
225
50
335.0
282
60
399.4
350
70
490.6
456
80
590.6
585
90
691.7
713
98
765.6
838
Table 3. API Gravity Assay.
Liq. vol. %
API gravity
13
63.28
33
54.86
57
45.91
74
38.21
91
26.01
Table 4. Characterisation of the process feed (Preheat Crude), basic temperature intervals.
Temperature Traditional method of characterisation New method of characterisation
range, oC 37.8 - 426.7
28 pseudo-components, uniformly
28 real components
426.7 - 648.9
8 pseudo-components, uniformly
8 pseudo-components, uniformly
648.9 - 760
2 pseudo-components, uniformly
2 pseudo-components, uniformly
Table 5. Resulting selection of real components (without the known light-end).
Number of No. Temp. interval [K]
Component
Tb
M
API
candidate
[K]
[kg/kmol]
grav.
components in interval
5
310.95 - 324.84
2,2-dimethylbutane
322.88
86.18
84.90
3
6
324.84 - 338.73
2-methylpentane
333.41
86.18
83.60
3
7
338.73 - 352.62
n-hexane
341.88
86.18
81.60
3
8
352.62 - 366.51
2-methylhexane
363.20
100.21
75.70
12
9
366.51 - 380.40
2,2-dimethylhexane
379.99
114.23
70.70
9
10
380.40 - 394.29
2-methylheptane
390.80
114.23
70.00
27
11
394.29 - 408.17
2,4-dimethylheptane
406.05
128.26
65.20
23
12
408.17 - 422.06
2,3-dimethylheptane
413.65
128.26
62.30
46
13
422.06 - 435.95
2,6-dimethyloctane
433.56
142.29
61.80
62
14
435.95 - 449.84
5-methylnonane
438.26
142.29
60.60
45
15
449.84 - 463.73
n-butylcyclohexane
454.13
140.27
44.70
25
16
463.73 - 477.62
n-undecane
469.04
156.31
58.60
12
17
477.62 - 491.51
n-dodecane
489.43
170.34
56.50
8
18
491.51 - 505.40
n-hexylbenzene
499.30
162.27
32.58
2
19
505.40 - 519.29
n-tridecane
508.58
184.37
54.60
8
20
519.29 - 533.18
n-tetradecane
526.66
198.38
53.60
6
21
533.18 - 547.07
n-pentadecane
543.77
212.41
51.80
14
22
547.07 - 560.96
n-hexadecane
559.94
226.43
50.60
7
23
560.96 - 574.85
n-decylbenzene
571.10
218.37
33.12
6
24
574.85 - 588.74
n-heptadecane
575.30
240.46
49.50
4
25
588.74 - 602.63
n-octadecane
589.86
254.48
48.60
4
26
602.62 - 616.51
n-nonadecane
603.80
268.51
47.80
8
27
616.51 - 630.40
n-uneicosane
629.65
296.56
46.29
5
28
630.40 - 644.29
n-dodecosane
641.76
310.59
45.74
3
29
644.29 - 658.18
n-tricosane
653.37
324.61
45.04
6
30
658.18 - 672.07
n-tetracosane
664.43
338.64
44.72
3
31
672.07 - 685.96
n-pentacosane
675.04
352.67
44.27
2
32
685.96 - 699.85
n-heptacosane
695.26
380.72
43.46
1
Table 6. Resulting substitute mixtures.
No.
Substitute mixture of pseudo-
Substitute mixture of real
components
components and pseudo-components
Real component /
Liq.vol. %
pseudo-component
Real component /
Liq.vol. %
pseudo-component
1
i-butane
0.19
i-butane
0.19
2
n-butane
0.11
n-butane
0.11
3
i-pentane
0.37
i-pentane
0.37
4
n-pentane
0.46
n-pentane
0.46
5
NBP_109
1.05
2,2-dimethylbutane
2.47
6
NBP_135
0.99
2-methylpentane
0.00
7
NBP_161
1.25
n-hexane
1.54
8
NBP_185
1.50
2-methylhexane
3.04
9
NBP_210
1.64
2,2-dimethylhexane
0.20
10
NBP_235
1.84
2-methylheptane
2.22
11
NBP_261
2.10
2,4-dimethylheptane
1.25
12
NBP_286
2.42
2,3-dimethylheptane
2.90
13
NBP_311
2.81
2,6-dimethyloctane
3.25
14
NBP_336
3.12
5-methylnonane
0.00
15
NBP_361
3.14
n-butylcyclohexane
5.17
16
NBP_386
3.14
n-undecane
2.64
17
NBP_411
3.08
n-dodecane
5.20
18
NBP_436
2.83
n-hexylbenzene
0.00
19
NBP_461
2.64
n-tridecane
3.23
20
NBP_486
2.55
n-tetradecane
3.67
21
NBP_511
2.47
n-pentadecane
2.45
22
NBP_536
2.41
n-hexadecane
3.30
23
NBP_561
2.36
n-decylbenzene
0.21
24
NBP_587
2.30
n-heptadecane
1.40
25
NBP_612
2.28
n-octadecane
2.65
26
NBP_637
2.31
n-nonadecane
2.77
27
NBP_662
2.30
n-uneicosane
4.54
28
NBP_687
2.22
n-dodecosane
0.00
29
NBP_712
2.09
n-tricosane
3.51
30
NBP_737
1.91
n-tetracosane
0.00
31
NBP_762
1.73
n-pentacosane
3.06
32
NBP_787
1.62
n-heptacosane
1.19
33
NBP_825
3.03
NBP_821
2.78
34
NBP_875
2.90
NBP_868
3.02
35
NBP_925
2.84
NBP_919
2.90
36
NBP_975
2.80
NBP_970
2.86
37
NBP_1025
2.76
NBP_1021
2.81
38
NBP_1075
2.73
NBP_1072
2.79
39
NBP_1125
2.71
NBP_1123
2.77
40
NBP_1175
2.73
NBP_1175
2.78
41
NBP_1251
5.65
NBP_1251
5.65
42
NBP_1372
8.64
NBP_1372
8.64
Appendix: Detailed Specifications for the Example
All plant data used for the simulation of the oil refining example originated from the HYSYS.Plant documentation (Hyprotech, 1998) and the overview is given by Tables A.1 - A.7.
Table A.1. Stream parameters for crude oil.
Temperature, oC
Pressure, kPa
232.2
517.1
Hot Crude
343.3
448.2
Tower Feed
338.8
448.2
Stream Preheated Crude (flowrate 662.4 m3/h)
Table A.2. Configuration of the atmospheric crude column.
Unit operation
Theoretical Stages
Main tray section
29
Condenser (for the main
1
tray section, partial) Kerosene side stripper
3
Reboiler for the kerosene
1
side stripper Diesel side stripper
3
AGO side stripper
3
Total
40
Table A.3. Connectivity for the Atmospheric crude column.
Stream
Description
From unit operation,
To unit operation, stage
stage TowerFeed
hot crude fed to the
-
main tray section, 28
-
main tray section, 29
column BottomSteam
steam fed to the bottom of the column
Naphtha
naphtha product
condenser
-
WasteH2O
waste water
condenser
-
OffGas
overhead vapor product
TrimDuty
energy stream - trim duty
-
main tray section, 28
Residue
crude atmospheric residue
main tray section, 29
-
Kerosene
kerosene product
reboiler for the
-
-
kerosene SS Diesel
diesel product
diesel SS, 3
-
AGO
atmospheric gas oil
AGO SS, 3
-
-
reboiler for the kerosene
product KeroSS_Energy
kerosene SS reboiler duty
SS KeroSS_Draw
liquid draw stream
main tray section, 9
kerosene SS, 1
KeroSS_Return
vapor return stream
kerosene SS, 1
main tray section, 8
DieselSS_Draw
liquid draw stream
main tray section, 17
diesel SS, 1
diesel SS, 1
main tray section, 16
DieselSS_Return vapor return stream
AGOSS_Draw
liquid draw stream
main tray section, 22
AGO SS, 1
AGOSS_Return
vapor return stream
AGO SS, 1
main tray section, 21
DieselSteam
steam fed to the bottom of
-
diesel SS, 3
-
AGO SS, 3
the diesel SS AGOSteam
steam fed to the bottom of the AGO SS
PA_1_Draw
draw for pump around 1
main tray section, 2
pump around cooler 1
PA_1_Return
return for pump around 1
pump around cooler 1
main tray section, 1
PA_1_Q
cooler 1 duty
pump around cooler 1
-
PA_2_Draw
draw for pump around 2
main tray section, 17
pump around cooler 2
PA_2_Return
return for pump around 2
pump around cooler 2
main tray section, 16
PA_2_Q
cooler 2 duty
pump around cooler 2
-
PA_3_Draw
draw for pump around 3
main tray section, 22
pump around cooler 3
PA_3_Return
return for pump around 3
pump around cooler 3
main tray section, 21
PA_3_Q
cooler 3 duty
pump around cooler 3
-
Table A.4. Parameters of column internal material streams.
Stream
Flowrate, m3/h
Naphtha
132.5
OffGas
0
Kerosene
86.12
Diesel
112.6
AGO
33.12
PA_1_Draw
331.2
PA_2_Draw
198.7
PA_2_Draw
198.7
Table A.5. Parameters of column steam supply.
Parameter
Stream Flowrate, kg/h
Temperature, oC
Pressure, kPa
BottomSteam
3402
190.6
1034.2
DieselSteam
1361
148.9
344.7
AGOSteam
1134
148.9
344.7
Table A.6. Column duty streams.
Stream
Duty, kJ/h
KeroSS_Energy
7.913×106
PA_1_Q
-5.803×107
PA_2_Q
-3.693×107
PA_3_Q
-3.693×107
Table A.7. Various column parameters.
Parameter
Value
Condenser pressure, kPa
135.8
Condenser pressure drop, kPa
62.1
Bottom stage pressure, kPa
225.5
Kerosen SS boil up ratio
0.75
Overflash specification, %
3.5
(or tray net liquid flow from stage 27)
(23.19 m3/h)
New Approach to the Characterization of Petroleum Mixtures Used in the Modelling of Separation Processes Egon Eckert and Tomáš Vaněk
Figures
Tb TI+1
Ti+1 Tbi Ti
T2
T1
0
xi
Φ
1
Figure 1: Definition of a pseudo-component and of its content in the substitute mixture from the TBP curve.
450 400
M [kg/kmol]
350 300 250 200 150 100 50 0
0
100
200
300
400
500
600
700
800
Tb [K] Figure 2. Plot of molecular weights vs. normal boiling points of hydrocarbon components from the HYSYS.Plant database.
350 300
API gravity
250 200 150 100 50 0 -50
0
100
200
300
400
500
600
700
800
Tb [K] Figure 3. Plot of API gravities vs. normal boiling points of hydrocarbon components from the HYSYS.Plant database.
Tb TI+1
Tbi
Ti+1 Ti T2
T1
0
Φ iL Φ i Φ iR
1
Φ
Figure 4. Situation after the selection of a real component in the ith primary temperature interval (Ti,Ti+1).
Tb
Tbi Tbi-1
0
Φ iL−1 Φ i −1 Φ iR−1 Φ iL Φ i Φ iR
1
Φ
Figure 5. Construction of intervals of fraction distilled in the second phase of the algorithm attached to real components selected in the first phase.
Figure 6. Simplified flowsheet for the crude oil processing example.
Figure 7. Subflowsheet for the atmospheric column.
400 350
M [kg/kmol]
300 250 200 150 100 50 300
350
400
450
500
550
600
650
700
Tb [K] Figure 8. Phase portrait for the molecular weight ( ) compared with values retrieved from the database for real components (£).
90
80
API gravity
70
60
50
40
30 300
350
400
450
500
550
600
650
700
Tb [K] Figure 9. Phase portrait for the API gravity ( ) compared with values retrieved from the database for real components (£).
1100 1000 900
Tb [K]
800 700 600 500 400 300 200 0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Φ Figure 10. Measured TBP curve (● ) compared with theoretical TBP curve for the substitute mixture gained by the new approach (£).
900
M [kg/kmol]
750
600
450
300
150
0 0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Φ Figure 11. Measured characterization curve for the molecular weight (● ) compared with the theoretical curve for substitute mixture gained by the new approach (£).
140 120
API gravity
100 80 60 40 20 0 0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Φ Figure 12. Measured characterization curve for the API gravity (● ) compared with the theoretical curve for substitute mixture gained by the new approach (£).
1100 1000 900
Tb [K]
800 700 600 500 400 300 200 0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Φ Figure 13. Comparison of theoretical TBP curves for fractionation products (bottom to top: Offgas, Naphtha, Kerosene, Diesel, AGO, Residue) employing the traditional (¢) and the new (£) approaches to characterization.
