Influence of thermodynamic-property models on the simulation of LNG evaporation and liquefaction processes

Florian Dauber Ruhr-University Bochum Germany F.Dauber@thermo.rub.de

Co-author Prof. Dr. Roland Span

Page 2 of 13

1.

ABSTRACT

Results of process simulations for liquefaction and vaporization processes in the natural gas industry are highly influenced by the thermodynamic-property model used. A higher accuracy for process modelling can be achieved by implementing new precise equations of state for natural gases and LNG into simulation tools. The GERG-2004 [1] is a wide-range equation of state, which is formulated for mixtures consisting of up to 18 specified components. It covers the gas and liquid phase as well as the supercritical region and the vapor-liquid equilibrium with the highest accuracy possible. The formulation was adopted as the new reference equation of state for natural gases and the development of a corresponding ISO standard is in process. However, so far the GERG-2004 is not used in commercial, LNG related simulation software. In order to implement the GERG-2004 into various simulation tools, the software available for GERG-2004 [2] has been adapted to the CAPE-OPEN standard. This standard defines rules and interfaces that allow CAPE (Computer-Aided Process Engineering) applications or components to interoperate. This contribution describes the way in which the GERG-2004 was integrated into common simulation tools. Compared to typical engineering equations of state, significant differences were found for the calculation of thermodynamic properties, e.g., for saturated liquid densities or in the two phase region. These differences result in significant changes for calculated heat exchanger surfaces in the corresponding stages when modelling LNG vaporization processes. Results found for the influence of thermodynamic-property models on the simulation of natural gas liquefaction processes are discussed by means of a model for the most often used technology in LNG production (C3MR).

Copyright 2008 IGRC2008

Page 3 of 13

CONTENTS

1. 2. 3.

Abstract ................................................................................................. 2 Introduction........................................................................................... 4 The GERG-2004 Equation of State for Natural Gases and Other Mixtures ....................................................................................... 4
Comparison of the GERG-2004 with Experimental Data and Values from Other Equations of State .............................................................................................5 Implementation of the GERG-2004 into simulation tools ..........................................6

3.1. 3.2.

4.
4.1. 4.2.

Influence of Property Models on the Simulation of LNG-Processes ........ 8

Simulation of a LNG Vaporization Process ..............................................................8 Simulation of a Natural Gas Liquefaction Process .................................................. 10

5. 6. 7.

Conclusion............................................................................................ 12 Acknowledgment ................................................................................. 12 References ........................................................................................... 13

Copyright 2008 IGRC2008

Page 4 of 13

2.

INTRODUCTION

To date natural gas as an energy source of the future is only getting to Germany via pipelines, which run through a multitude of states. In other industrialized countries the gas supply is at least partly based on Liquefied Natural Gas (LNG). Most likely the German supply system will develop in a similar way the importance of LNG and of related technologies increases even further. Evaporation and liquefaction are essential processes in the LNG process chain. High expectations regarding economic efficiency, product quality and environmental safety lead to increasing demands on these processes. This applies to both the design and the operation of a specific system, e.g. to the optimal use of resources while vaporizing LNG. Fundamental contributions to optimization and operation of a system are expected from the simulation of processes. Therefore, an accurate representation of thermophysical properties and energy balances of the simulated process is essential. Results of process simulations are highly influenced by the thermodynamicproperty model used. New highly accurate equations of state for natural gas and LNG represent an unused potential for accurate process modelling.

3.

THE GERG-2004 EQUATION OF STATE FOR NATURAL GASES AND OTHER MIXTURES

The GERG-2004 is an equation of state that describes mixtures of arbitrary composition consisting of up to 18 specified components. The equation is valid in the gas phase, in the liquid phase, in the supercritical region and for the vapour-liquid equilibrium. Comparisons with experimental data for natural gases show that for the calculation of thermodynamic properties a higher accuracy can be achieved by the GERG-2004 than with the established international standard, the AGA8-DC92 equation of state. The new equation stands out due to the fact that for a wide range of validity thermal and caloric properties can be calculated with the highest accuracy possible, which enables its application to the layout of LNG processes. The GERG-2004 was adopted by the GERG (Groupe Europen de Recherche Gazires) as the new reference equation of state for natural gases. The development of a corresponding ISO standard is proceeding. The GERG-2004 formulation is based on a multi-fluid approximation, which is explicit in the reduced Helmholtz energy = a / (RT) depending on the density , the temperature T, and the composition x (mole fraction) of the mixture.
.

Copyright 2008 IGRC2008

Page 5 of 13

3.1.

Comparison of the GERG-2004 with Experimental Data and Values from Other Equations of State

The representation of experimental data by the GERG-2004 as well as the comparison with other property models proves that high accuracies can be achieved for the calculation of thermodynamic properties of LNG mixtures. Extensive comparisons for a multitude of properties were given by Kunz et al. [1]. In this article we restrict ourselves to comparisons to data for LNG densities. The experimental data were taken from Haynes (1982) [3] and Hiza and Haynes (1980) [4] and were determined at the National Institute of Standards and Technology (NIST, former NBS) in Boulder, Colorado, USA. The measured data describes LNG densities on the saturated liquid line. The uncertainty specified by the authors is / 0.1%. Table 3.1 gives the LNG compositions for which the densities were experimentally determined.
Tab 3.1 Molar composition of the gases used for density measurements by Haynes (1982) [4] and Hiza and Haynes (1980) [3]; these data are used for the comparison of property models.

Author Haynes (1982) Hiza and Haynes (1980) Haynes (1982) Haynes (1982) Haynes (1982) Haynes (1982) Hiza and Haynes (1980) CH4 85.89 85.44 84.55 90.61 88.22 85.93 81.30 N2 0.60 0.97 1.38 4.25

Composition [mol%] C2H6 11.53 5.042 8.153 6.026 7.259 8.477 4.750 C3H8 1.341 4.038 4.778 2.154 2.561 2.980 4.870 n-C4H10 0.705 2.901 1.252 0.306 0.492 0.707 2.420 i-C4H10 0.530 2.577 1.259 0.300 0.490 0.519 2.410

The calculated percentage deviations of values calculated with the Peng-Robinson, revised Klosek-McKinley and GERG-2004 equations from measured saturated liquid densities are presented in Figs. 3.1 to 3.3. The cubic equation of Peng and Robinson (1976) [5] and the revised Klosek-McKinley equation of McCarty (1980) [6] are among the methods mostly used when calculating LNG densities. Values calculated with the revised Klosek-McKinley equation deviate from the 27 measured values by (0.00040.19)%. This excellent agreement is explained by the fact that the equation has been fitted to the measured densities. However, the revised Klosek-McKinley equation describes only saturated-liquid densities. It does not consider the pressure dependence of thermodynamic properties and does not permit calculations of caloric properties. Thus, the revised Klosek-McKinley equation is unsuitable for process simulation. The values calculated using the GERG-2004 equation deviate from the experimental data by (0.060.32)%. This result agrees with the uncertainty of / (0.10.5)% specified by the authors of the GERG-2004. The densities calculated with the Peng-Robinson equation deviate by up to -12%.
0,5 100 ' / '

-0,5 100
Fig. 3.1

105

110

115 120 Temperatur [K]

125

130

135

Percentage deviations of measured saturated liquid densities for selected LNG compositions from corresponding values calculated with the GERG-2004 equation.

Copyright 2008 IGRC2008

Page 6 of 13
0,5 100 ' / '

-0,5 100
Fig. 3.2

105

110

115 120 Temperatur [K]

125

130

135

Percentage deviations of measured saturated liquid densities for selected LNG compositions from corresponding values calculated with the revised Klosek-McKinley equation.

15 100 ' / '

-15 100
Fig. 3.3

105

110

115 120 Temperatur [K]

125

130

135

Percentage deviations of measured saturated liquid densities for selected LNG compositions from corresponding values calculated with the Peng-Robinson equation.

Just as an example, this comparison underlines that the GERG-2004 equation of state is the most accurate property model available for the simulation of LNG processes to date. 3.2. Implementation of the GERG-2004 into simulation tools

As shown above, the GERG-2004 as the new reference equation of state for natural gases has a high potential for accurate process-modelling. So far the GERG2004 is not used in any commercial simulation software. In order to implement a GERG-2004 Dynamic Link Library (DLL) [2] into various simulation tools, a thermodynamic property interface has been designed meeting the CAPE-OPEN standard. This open standard defines rules and interfaces that allow CAPE (Computer-Aided Process Engineering) applications or components to interoperate. Due to the fact that this standard is multi-platform and supported by most of the commonly used simulation tools, it enables software components supplied by third parties, such as physical property packages or unit operation models, to be used in plug and play mode within commercial process modelling software tools. In this context a property package describes a software component which is both a thermodynamic property calculator and an equilibrium calculator for mixtures containing a specific set of compounds occurring in a specific number of physical states. The successful development of the GERG-2004 property package enables the application of the GERG-2004 model in a process modelling environment. The property package is limited only to the range of validity of the GERG-2004, which covers temperatures in the range of 75 K T 700 K and pressures of p 70 MPa for mixtures containing up to 18 of the most relevant natural gas components. For process simulations an accurate determination of the phases in a natural gas mixture or a pure fluid is required. Furthermore, different combinations of independent variables are required to calculate thermodynamic properties in a process
Copyright 2008 IGRC2008

Page 7 of 13

modelling environment. As the GERG-2004 DLL only supports property calculation with temperature and pressure or temperature and density as independent variables, algorithms for additional combinations of independent variables had to be developed for the GERG-2004 property package. In order to allow property calculations with all combinations of independent variables needed for process simulations, iterative routines have been written using the bisection method. The iterations are based on values calculated by the GERG-2004. Thus, the accuracy of the model is not affected; it is independent of the chosen set of independent variables. The different combination of independent variables supported by the GERG-2004 DLL and the GERG-2004 property package are shown in Tab. 3.2.
Tab 3.2 Combinations of independent variables supported by the GERG-2004 DLL and the GERG-2004 property package for calculating thermodynamic properties of a pure component or a mixture.

GERG-2004 property package Temperature - Pressure Temperature - Enthalpy Temperature - Entropy Temperature - Vapor Fraction Pressure - Enthalpy Pressure - Entropy Pressure - Vapor Fraction

GERG-2004 DLL Temperature - Pressure Temperature - Density

The thermodynamic properties that can be directly calculated with the GERG-2004 property package in a process modelling environment are shown in Tab. 3.3.
Tab 3.3 Thermodynamic properties that can be directly calculated with the embedded GERG-2004 property package in a process modelling environment.

Temperature Pressure Density Enthalpy Entropy Isobaric heat capacity Gibbs free energy Compressibility factor

T p h s cp g Z

Fugacity coefficient Derivative of the fugacity coefficients with regard to temperature Derivative of the fugacity coefficients with regard to pressure Logarithm of the derivative of the fugacity coefficients with regard to temperature Logarithm of the derivative of the fugacity coefficients with regard to pressure Logarithm of the K-factors

i
i T i p
ln

i T
i p

ln

lnKi

Until now the property package has been tested successfully in simulation tools like Aspen Plus, Aspen HYSYS, CHEMCAD, Pro/II and UniSim.

Copyright 2008 IGRC2008

Page 8 of 13

4.

INFLUENCE OF PROPERTY MODELS ON THE SIMULATION OF LNG-PROCESSES

The following chapter examines the influence of thermodynamic-property models on the simulation of LNG evaporation and liquefaction processes with Aspen Plus and Aspen Hysys, the most commonly used simulation tools in the field of process engineering. 4.1. Simulation of a LNG Vaporization Process

In the following the influence of thermodynamic-property models on the simulation results obtained for vaporization of liquefied natural gas across the twophase area (vapor/liquid) or at a supercritical pressure is discussed. The simulation tool Aspen Plus is used to simulate these vaporization processes. The thermodynamicproperty models examined are the cubic Peng-Robinson and Redlich-Kwong-Soave equations and the GERG-2004 reference equation of state. A typical effect in such comparisons is that relative deviations are much smaller, if the evaporation process is modeled in a single step, since positive and negative deviations compensate for each other. This is why the impact of property models is frequently underestimated. For the detailed design of, e.g., a heat exchanger stepwise calculations are required.

Fig. 4.1

Screenshot of a modeled LNG multi-stage vaporization within Aspen Plus.

Fig. 4.1 illustrates a simulated LNG vaporization set-up. In order to show the influence of property models on lay-out data of LNG heat exchangers, the temperature of the liquefied natural gas to be vaporized is increased by 10 K downstream of each heat exchanger. For this purpose, water with a temperature of 285 K is sent through a counter-current heat exchanger. In this way, the natural gas is heated from a temperature of 100 K to 280 K at a supercritical pressure of 9.75 MPa. Fig. 4.2 shows deviations in heat exchanger surfaces calculated on the basis of the cubic PengRobinson and Redlich-Kwong-Soave equations from the corresponding values obtained with the GERG-2004 property package.
Copyright 2008 IGRC2008

Page 9 of 13
10 100 * A / A 5 0 -5 -10 100 120 140 160 180 200 220 240 260 280 Temperature / K

Fig. 4.2

Percentage deviations of the heat exchanger surfaces calculated on the basis of the cubic Peng-Robinson ( ) and Redlich-Kwong-Soave ( ) equations from the heat exchanger surfaces calculated on the basis of the GERG-2004 property package (baseline) for a simulated multi-stage LNG vaporization process at a supercritical pressure of 9.75 MPa.

Calculations on the basis of the Peng-Robinson equation result in a maximum deviation of -9.3% from the corresponding values determined with the GERG-2004 property package. The deviations are particularly high in the range above the critical point. Maximum deviations of 7.5% were found for heat exchanger surfaces in the simulation based on the Redlich-Kwong-Soave equation. By heating up liquefied natural gas across the two-phase area (vapor/liquid) at ambient pressure (0.101325 MPa) comparisons for a LNG flash vaporization can be made. Therefore a certain amount of energy (1,000 kJ/kmol) is added in each step. Thus, the enthalpy of the LNG mixture increases. Fig. 4.3 demonstrates the deviations of the vapor fractions calculated by the cubic Peng-Robinson and Redlich-KwongSoave equations from the corresponding values determined by the GERG-2004 property package.
100 (x cub. Gl. - x GERG ) / x cub. Gl. 30 15 0 -15 -30 -16000 -14000 -12000 -10000 Enthalpy [kJ/kmol] -8000 -6000 -4000

Fig. 4.3

Absolute deviations of the vapor fractions calculated on the basis of the cubic equations of Peng-Robinson ( ) and Redlich-Kwong-Soave ( ) from GERG-2004 property package (baseline) for a simulated LNG flash vaporization process at a pressure of 0.101325 MPa. The dashed lines represent the phase boundaries.

Deviations between the calculated vapor fractions are rather high. The largest deviations are found particularly near the saturated liquid line where the low-boiling constituents vaporize. These constituents, like methane, make up the main share in the LNG mixture. Thus, the vapor fraction rises rapidly here. Different descriptions of the phase equilibrium by the different equations of state lead to different compositions of the liquid and vapor phases. The largest deviations from the GERG-2004 property package are found for the Redlich-Kwong-Soave equation. They are as high as -28.9% downstream of the first heat exchanger. The maximum deviation for calculations on the basis of the Peng-Robinson equation is 5.4%.
Copyright 2008 IGRC2008

Page 10 of 13

4.2.

Simulation of a Natural Gas Liquefaction Process

Before natural gas is shipped it has to be liquefied by cooling it down to a temperature of around 110 K at ambient pressure. For more than 30 years the majority of the worlds LNG is produced using a technology called C3MR, where C3 stands for propane as the precooling fluid and MR for mixed refrigerant as the liquefaction and sub-cooling fluid. The technology developed by Air Products has the highest efficiency on the market. Currently a plant using this technology is built up in Peru for example. The C3MR process is modeled with Aspen HYSYS utilizing the GERG2004 property package. In the following section the C3MR process is described and the results are compared with those achieved when simulating the process using the frequently used model by Kesler, Lee and Plcker [7, 8]. After the acid gas is removed the natural gas enters the C3MR process and is dried and pre-cooled to about 241 K by propane. The pre-cooled gas is now liquefied and sub-cooled to around 123 K in the main cryogenic heat exchanger by a mixed refrigerant flowing down on the shell side. In order to precool the natural gas, propane is compressed to a high pressure. That way it can be condensed by ambient air or cooling water. By expanding the propane in several stages the temperature is reduced, allowing the propane to chill the natural gas. The propane also precools the compressed mixed refrigerant. After being precooled the mixed refrigerant is separated within a high pressure separator. In the main cryogenic heat exchanger the vapor and the liquid streams are cooled, liquefied and sub-cooled. The sub-cooled streams are throttled to further reduce the temperature in order to liquefy and subcool the natural gas. The LNG leaves the process with a temperature of around 112 K at ambient pressure. [9] A flow sheet of the C3MR process as it is modeled in Aspen HYSYS is shown in Fig. 4.4.

Fig. 4.4

Aspen HYSYS screenshot of a C3MR liquefaction process for natural gas.

Copyright 2008 IGRC2008

Page 11 of 13

To show the influence of property models on the C3MR liquefaction process, the results found with the GERG-2004 property package are compared with those found using the Lee-Kesler-Plcker equation. The most crucial deviations can be found for the high pressure separator where the vapor and the liquid phase of the mixed refrigerant are separated as illustrated in Fig. 4.5.

Fig. 4.5

Detail of a modeled C3MR liquefaction process for natural gas showing the high pressure separator for the mixed refrigerant.

Deviations of 10% can be found for the vapor fraction of the mixed refrigerant. The separated vapor and the liquid stream differ in their mass flow by up to 14%. A difference of 2.2% can be stated for the calculated energy demand of the mixed refrigerant compressors. Tab. 4.1 presents the worksheets of streams entering and leaving the high pressure separator and points out the results of the comparison between the GERG-2004 property package and the Lee-Kesler-Plcker model.
Tab 4.1 Worksheet for the streams entering and leaving the high pressure separator of a modeled C3MR liquefaction process.

GERG-2004 property package MR 0.48 0.03 0.40 0.06 0.03 0.3385 CH4 N2 C2 H6 C3 H8 i-C4H10 MR-V 0.7101 0.0646 0.2110 0.0117 0.0027 0.1718 CH4 N2 C2 H6 C3 H8 i-C4H10 MR-L 0.3623 0.0123 0.4967 0.0847 0.0440 0.4547

Lee-KeslerPlcker model 0.48 0.03 0.40 0.06 0.03 0.3072 0.7222 0.0684 0.1968 0.0104 0.0022 0.1508 0.3726 0.0130 0.4901 0.0820 0.0423 0.4616

Deviations [%]

Composition [mol-%]

CH4 N2 C2 H6 C3 H8 i-C4H10

Vapor fraction Composition [mol-%]

-10.1927 1.6741 5.6078 -7.2277 -12.2187 -19.0619 -13.8793 2.7796 5.0862 -1.3480 -3.3329 -3.9689 1.5022

Mass flow [kg/s] Composition [mol-%]

Mass flow [kg/s] Copyright 2008 IGRC2008

Page 12 of 13

These differences are believed to result from a less accurate representation of thermodynamic properties and particularly of properties at vapor-liquid equilibrium by the Lee-Kesler-Plcker equation. Deviations on this order of magnitude result in far sub-optimal plant layout.

5.

CONCLUSION

The GERG-2004 is an equation of state that covers the whole fluid region for different natural gases and other mixtures with the highest accuracy possible. This reference equation of state has been made accessible for accurate LNG processmodelling by developing the GERG-2004 property package. The thermodynamic property package meets the CAPE-OPEN standard which enables the implementation into most of the commonly used simulation tools and moreover assures stable and consistent calculations. The influence of thermodynamic-property models of different qualities regarding simulation of processes in the LNG process chain has been investigated. For example, modelling the surface size of heat exchangers for LNG vaporization on the basis of the Peng-Robinson equation results in a maximum deviation of -9.3% from the corresponding values determined with the highly accurate GERG-2004 property package. Comparisons for the influence of different property models on a model for an industrially used liquefaction process show deviations of up to 14% for mass flows. Deviations on this order of magnitude should not be disregarded as they may lead to sub-optimal solutions in terms of technology and economics.

6.

ACKNOWLEDGMENT

This work is carried out as part of the "competence centre thermodynamics of gases" initiative jointly financed by e.on Ruhrgas and by the Ministry of Innovation, Science, Research and Technology of North Rhine-Westphalia.

Copyright 2008 IGRC2008

Page 13 of 13

7.
[1]

REFERENCES
Kunz, O.; Klimeck, R.; Wagner, W.; Jaeschke. M. (2007): The GERG-2004 WideRange Reference Equation of State for Natural Gases and Other Mixtures. GERG Technical Monograph 15. Fortschr.-Ber. VDI, Reihe 6, Nr. 557, VDI Verlag, Dsseldorf. Kunz, O.; Wagner, W. (2006): Software package for the calculation of thermodynamic properties from the GERG-2004 wide-range reference equation of state for natural gases and other mixtures. Lehrstuhl fr Thermodynamik, RuhrUniversitt Bochum. Hiza, M. J.; Haynes, W. M. (1980): Orthobatic liquid densities and excess volumes for multicomponent mixtures of low molar-mass alkanes and nitrogen between 105 and 125K. J. Chem. Thermodynamics, 12, 1 10. Haynes, W. M. (1982): Measurements of orthobatic-liquid densities of multicomponent mixtures of LNG components (N2, CH4, C2H6, C3H8, CH3CH(CH3)CH3, C4H10, H3CH(CH3)C2H5, and C5H12) between 110 and 130 K. J. Chem. Thermodynamics, 14, 603 612. Peng, D. J.; Robinson, D. B. (1976): A New Two-Constant Equation-of-state. Ind. Eng. Chem. Fundam., Vol. 15, 59 64. McCarty, R. D. (1980): Four Mathematical Models for the Prediction of LNG Densities. National Bureau of Standards Technical Note 1030, Boulder, Colorado, USA. Lee, B. I.; Kesler, M. G. (1975): A generalized thermodynamic correlation based on threeparameter corresponding states. AIChE J., 21, 510 527. Plcker, U.; Kamp, H; Prausnitz, J. (1978): Calculation of high-pressure vaporliquid equilibria from a corresponding-states correlation with emphasis on asymmetric mixtures. I&EC Proc. Des. & Dev. 17, 324 332. Pillarella, M.; Liu, Y.-N.; Petrowski, J.; Bower, R. (2007): The C3MR Liquefaction Cycle: Versatility for a Fast Growing Ever Changing LNG Industry. Air Products, Paper PS2-5.

[2]

[3]

[4]

[5]

[6]

[7]

[8]

[9]