Accurate Prediction of Compressibility

~actors by the GERG Virial Equation

Manfred Jaeschke, Ruhrgas AG; Serge Audibert, Gaz de France; Patricia van Caneghem, Distrigaz;
A.E. Humphreys, British Gas; Renee Janssen-van Rosmalen, SPE, Gasunie; Quinto Pellei, SNAM;
and J.A. Schouten and J.P.J. Michels, Van der Waals Laboratory

Summary. The European Gas Research Group (GERG, Groupe Europeen de Recherches Gazieres) performed an extensive research
project involving the measurement of compressibility factors of pure gases and binary mixtures and of natural gases in the temperature
range of 265 to 335 K and at pressures up to 12 MPa. These pure gas and binary mixture data, together with high-quality data from
the literature, were used to develop the GERG virial equation for the accurate prediction of the compressibility factor of natural gas
(multicomponent) mixtures. Pressure, temperature, and a 13-component composition are used as input data. Eighty-four sets of ex-
perimental natural gas data made up of more than 4,000 data points were used to validate the GERG virial equation of state (EOS).
The target accuracy of the GERG virial equation is ±0.1 %. A comparison of experimental and predicted compressibility factors shows
that this target was achieved. The average root-mean-square (RMS) error for the differences between the experimental and the predicted
compressibility factors is 0.06% for pressures up to 12 MPa. Even in the high-pressure range between 8 and 12 MPa, the RMS error
is only 0.07%.

Introduction
Metering of large-volume flows of natural gas at high pressure re-
quires accurate knowledge of the compressibility factor. A num-
ber of gas companies use modified versions of the American Gas
Assn. (AGA) NX-19 equation. 1,2 These methods are satisfactory
when the pressure is not too high and the temperature range is limit-
ed. In certain cases, however, the difference between measured and
predicted compressibility factors may be as great as 1 %.
British Gas used an extended corresponding-states method, and
Gaz de France used a modified version of the Redlich-Kwong EOS.
These two equations show RMS errors of about 0.25 % between
measured and predicted data from an early (1982) version of the
GERG data banle 3 None of these methods mentioned is fully satis-
factory. Therefore, several years ago, GERG decided to develop
an EOS to predict the compressibility factor of natural gases with
an uncertainty of less than 0.1 %. The initial results of this work
were presented in 1986. 4 This paper presents details of the final
development of the GERG virial equation. The accuracy of the com-
pressibility factor calculation obtained is discussed.
Specifications of GERG Viria. Equation
The input data for calculation of the compressibility factor of natu-
ral gas by the GERG virial equation are temperature, pressure, and
composition of the mixture. For the composition of the mixture,
the mole fraction of each of the 13 groups of components listed
in Table 1 is required. Table 1 also gives the specifications for the
GERG virial EOS.
Presentation of GERG Viria. Equation
The behavior of a real gas may be written as
z=p/PmRT. ....................................... (1)
Compressibility factor, z, is a convenient measure of the devia-
tion of the behavior of a real gas from ideal (z = 1). As for any ideal
gas, z= I at zero pressure. At low temperature and pressure, gener-
ally z < 1 but may be > 1 at high pressure and temperature.
The virial EOS is a straightforward way to describe the behavior
of a real gas. This equation, truncated after the third term, has the
form
z=l+BM(T)Pm+CM(T)p;" ......................... (2)
where BM(T) and CM(T) are the second and third virial coeffi-
cients, which depend on temperature and gas composition. For a
gas mixture, the coefficients are determined by the mixing rule,
n
I: xixjBij(T) ...................... (3)
j =1
Copyright 1991 Society of Petroleum Engineers
n n n
and CM(T)= I: I: I: XiXjXkCijk(T), ............ (4)
i=1 j=1 k=1
where Xi' Xj, and xk represent the mole fractions of the ith, jth,
and kth component, respectively.
The virial coefficients of a mixture are thus obtained from the
virial coefficients of the pure components (Bii,C ii ;) and unlike-
interaction virial coefficients (B ij' C ij k)' The unlike-interaction viri-
al coefficients are independent of composition. Second-order poly-
nomials may be used to describe the temperature dependence of
all relevant virial coefficients:
Bij (T) =b~~) +b~y T +b~]> T2 ......................... (5)
and Cijk(T)=C~%+cmT +c~71T2 . ...................... (6)
It has been shown empirically5 that this simple quadratic tem-
perature dependence is sufficient to describe adequately the avail-
able good-quality data over the entire temperature range of interest.
The GERG virial equation was developed from Eqs. 2 through 6.
The virial coefficients were obtained from experimental data on
pure gases and binary and ternary mixtures. The data are stored
in the GERG data bank (Table 2). The work was carried out by
the van der Waals Inst. in Amsterdam under contract with
GERG.5,6
As a first step, second and third virial coefficients and the pa-
rameters of their temperature dependence were evaluated for the
pure gases from experimental PVT data. The unlike-interaction viri-
al coefficients were then calculated from the relevant binary mix-
ture data by subtraction of the now-known pure-gas contribution-
i.e., by rewriting Eqs. 3 and 4 for binary mixtures:
BM(T)=xt Bll +2x 1x 2B I2 +Xi B 22 ................... (7)
and CM(T)=x[C lll +3XrX2CJ12+3xIXiC122+xic222' .. (8)
The three unlike-interaction coefficients-Bu, C 1l2 , and C 122 -
were also assigned a quadratic temperature dependence (Eqs. 5 and
6). However, the number of experimental data or the measurement
accuracy often is insufficient to obtain adequately reliable values
for all these coefficients. Thus, the number of unknown variables
had to be reduced by use of an additional relation between the third
unlike-interaction virial coefficients,7 C ll2 and C 122:
ClJ2 -C m = V3(C lll -C222 ) . ....................... (9)
For similar reasons, the unlike triple-interaction coefficients can-
not reliably be obtained directly; instead, a combining rule is used:
Cijk =~ijk(CiiiCjjj C kkk ) 'h, . . . . . . . . . . . . . . . . . . . . . . . . . (10)
where usually ~ ij k = l.

SPE Production Engineering, August 1991 343

 TABLE 1-SPECIFICATIONS OF THE 1.400 . , . . . - - - - - - - - - - - - - - - - - ,
GERG VI RIAL EQUATION
all exp. data (4486)
(12 outside! 0.25%)
Specified maximum uncertainty, % ±0.1 ! 1,200
Temperature conditions, K 265~T~335 c=
CII
Pressure conditions, MPa 0~p~12
Gas composition, mol fraction e 1,000
'-
CII
c..
X
800

-
CH 4 ;::0.5 N2 ~0.5 CO 2 ~0.3 CII
C 2 H 6 ~0.2 C3He ~0.05 H2 ~0.1
C 4H 10 ~ 0.015 CO~0.03 He~0.005 o
600
'-
C S H'2 +benzene~0.005 CII
~
C 6H 14 + ethylbenzene~0.001
C 7 H 16 +toluene~0.001 5 400
c=
C eH,e +C 9 H 20 +ClOH22~0.001
200

TABLE 2-S0URCES OF EXPERIMENTAL DATA· o -0.2 -0.1 o 0.1 0.2

Data Sets percentage relative deviation
Pure Gases Binary Mixtures Ternary Mixtures
British Gas 2 (26) 3 (53) 1 (9) Fig. 1-Relative error distribution for all experimental z data
Distrigaz 0(0) 1 (20) 1 (17) of natural gases.
Gasunie 2 (188) 19 (717) 8 (196)
Gaz de France 0(0) 3 (53) 1 (14)
Ruhrgas AG 12 (1,091) 62 (4,324) 5 (332)
SNAM 0(0) 1 (12) 1 (12)
Literature 20 (1,069) 18 (768) 1 (40) temperature intervals
~ [265K. 275K)
Total 36 (2,374) 107 (5,847) 18 (620)
----- [275K. 285K)
• Number of data points given in parentheses. ~ 0.10
. -0-0- [285K. 295K)
c... ............... [295K. 315K)
o
c...
c...
- - - D15K. 335K]
~
LA
TABLE 3-UNCERTAINTY OF DIFFERENT TECHNIQUES CII
I
FOR MEASUREMENT OF PURE-GAS VI
COMPRESSIBILITY FACTOR (%) E
c...
~
p~8 MPa 8~p~12 MPa 0.05
Burnett apparatus 0.05 0.1
~

Grating interferometer 0.06 0.1

DEH z meter 0.1

Experimental Data and Numerical Values

of Virial Coefficients
The data used to evaluate the coefficients and to test the GERG
virial equation were obtained by two different techniques. First,
an expansion technique, e.g., the Burnett apparatus or the Des-
granges et Huot (DEH) z-meter, was used. 3,8,9 In this method, a
gas sample is expanded at constant temperature and z is deduced
from the measured change in pressure. Second, an optical method
was used, where the compressibility factor can be obtained through
measurement of the refractive index of a gas with a grating inter-
ferometer. IO ,1l The refractive index is linked through the Lorentz-
Lorenz law to the molar density and hence the compressibility fac-
tor of the gas investigated.
The measurement uncertainties depend on the pressure range (see
Table 3) and increase with pressure. When measuring compressi-
bility factors of gas mixtures or natural gases, we must acknowledge
an additional uncertainty related to uncertainties in the gas compo-
sition. The first criterion for accepting data in this correlation work
was the uncertainty of measured compressibility factor as estimat-
ed or quoted by the authors. Other considerations are that the pres-
sure and temperature ranges of the data cover the entire range of
application and that the data must be internally consistent. The proce-
dure of evaluating all the second and third virial coefficients is de-
scribed in more detail in Ref. 12. Table 4 gives the values for the
coefficients of the temperature expansion for the second and third
virial coefficients obtained in this correlation work. All interactions
for which no values are listed in Table 4 are ignored.
--
0.00
o 4 8 12
pressure p, MPa
Fig. 2-RMS error of all experimental z data (N =4,463) for
different p- T domains.
Performance of Truncated GERG Virial Equation
Test data. Twenty data sets for synthetic gas mixtures and 84 data
sets for natural gas that were not used in deriving the GERG virial
equation are available. 13 Table 5 lists the source of these data sets
and the number of data points.
Nederlandse Gasunie and Ruhrgas measured most of the natural
gas compressibility factors stored in the GERG data bank. Ruhr-
gas was the only company able to make measurement at pressures
up to 12 MPa. The majority of the data, for synthetic gas mixtures
containing at least four different components, were provided by
Gaz de France. The specified uncertainty in compressibility fac-
tors is always within ±O.1 %.13
The quality of the GERG virial equation was assessed by com-
paring the predicted z values with the experimental data in the GERG
data bank (obtained from DEH z meter from 0 to 8 MPa, Burnett
apparatus, optical measurements from 0 to 12 MPa, and literature
from 0 to 9 MPa).

344 SPE Production Engineering, August 1991

 TABLE 4-TEMPERATURE EXPANSION COEFFICIENTS

Second Virial Coefficients

8=b(O) +b(1)T +b(2)T2
(8 is in m 3 /kmol; T is in Kelvin)
Pure Gases/Mixtures b(O) b(1) b(2)

CH 4 - 2.98675 x 10 -1 1.33425 x 10- 3 -1.59761 x 10- 6

N2 -1.44600xlO- 1 7.40910 x 10- 4 - 9.11950 x 10- 7
CO 2 - 8.68340 x 10- 1 4.03760 x 10- 3 - 5.16570 x 10- 6
C 2H 6 -1.07320 4.64810 x 10- 3 - 5.60520 x 10- 6
H2 -1.10596xlO- 3 8.13385x10- 5 - 9.87220 x lO- s
C3 Ha -2.59920 1.19650 x 10- 2 -1.5291Ox10- 5
CO -1.30820 x 10- 1 6.02540 x 10- 4 - 6.44300 x 10- 7
C 4H lO - 7.08016 3.63100x 10- 2 - 5.03829 x 10- 5
He 2.06740 x 10- 2 - 5.13060 x 10- 5 7.24000x10- s
C 5H12 -1.11580x101 5.35740 x 10- 2 - 6.84970 x 10- 5
CH 4/N 2 -2.13606 x 10- 1 1.04585 x 10- 3 -1.31159x10- 6
CH 4 /C0 2 -3.56120x 10- 1 1.44963 x 10- 3 -1.58061 x 10- 6
CH 4/C 2H 6 - 4.99337 x 10 -1 2.05755 x 10- 3 - 2.30786 x 10- 6
CH 4/H 2 - 3.28913 x 10 -2 1.58947 x 10- 4 -9.16521 x 10- s
CH 4/C 3 H s -5.54110x 10- 1 1.85897x10- 3 -1.47793x10- 6
CH 4 /CO - 6.87290 x 10- 2 - 2.39381 x 10- 6 5.18195 x 10- 7
CH 4/C 4H lO -1.38708 6.89575 x 10- 3 -9.40171xlO- 6
CH 4/C 5 H 12 - 7.71367 x 10- 1 1.43827 x 10 - 3 1.25744x10- 6
CH 4/C 6H 14 -7.75140x10- 1 1.84050x10- 3 -4.00000 x 10- 7
CH 4/C 7 H 16 -2.09989 9.79707 x 10- 3 - 1.25373 x 10 - 5
CH 4/C aH 1a -2.39409 1.12199x 10- 2 -1.43976 x 10- 5
N 2 /C0 2 -3.39693 x 10- 1 1.61176x10- 3 - 2.04429 x 10- 6
N 2/C 2H 6 - 2.63553 x 10 - 1 1.06233 x 10- 3 -1.10155 x 10- 6
N2/H2 1.84506 x 10 - 2 -1.11895 x 10- 4 3.03122x10- 7
N2/C 3H S -4.31268 x 10- 1 1.84597 x 10- 3 - 2.08060 x 10- 6
N 2 /CO -1.22189x10- 1 5.21240 x 10- 4 - 4.37181 x 10- 7
N 2/C 4 H lO 9.07220 x 10- 2 -1.98016x10- 3 4.56313x10- 6
C0 2 /C 2 H 6 -1.12000 5.76913xlO- 3 - 8.13744 x 10- 6
C0 2 /H 2 - 7.57226 x 10- 2 3.56052 x 10- 4 - 3.57678 x 10- 7
C0 2 /C 3 H a -1.26372 5.84661 x 10- 3 -7.47500 x 10- 6
C0 2 /C 4 H 10 -1.88108 9.26431 x 10- 3 -1.24525 x 10- 5
C0 2 /C 5 H 12 -2.02135 9.54081 x 10- 3 -1.21639 x 10- 5
C 2H 6 /H 2 -1.74834x10- 1 1.03728 x 10- 3 -1.50332 x 10- 6
C2H6/C3Ha -1.42020 5.75230 x 10- 3 - 6.40000 x 10- 6
C 2H 6/C 4 H lO -2.83644 1.36800x10- 2 -1.81244 x 10- 5
C2H6/C5H12 -3.31426 1.52019 x 10- 2 -1.89007 x 10- 5
C3Ha/C4HlO -4.35393 2.13786 x 10- 2 - 2.87483 x 10- 5

temperature intervals 200

-0-0- [265K, 275K) all ex p. da taw i t h XN2 ~ 0.095
~ [275K, 285K) ." (625 )
~ 0.10
. -0-0- [285K, 295K)
+-
c:
.... ~ [295K,315K)
QI
e
o
....
.... - - - [315K, 335K] ....
QI
GI c..
I X

-
." QI
e
.... 100
0
0.05 ....
-0. I QI
~
e
:::I
c:

0.00 -I---_-~-~-~-_-_.,.....J o~~~~~~~~~~~~~

o 4 8 12 -0.2 -0.1 0 0.1 0.2

pressure p,MPa percentage relative deviation

Fig. 3-RMS error of all experimental z data measured with Fig. 4-Relative error distribution of the experimental z data
the grating interferometer or Burnett apparatus for different for natural gases containing more than 9.5% nitrogen and no
p- T domains. hydrogen.

SPE Production Engineering. August 1991 345

 TABLE 4-TEMPERATURE EXPANSION COEFFICIENTS (continued)

Third Virial Coefficients

C=c(O) +c(1)T +c(2)T2

(C is in m S/kmol 2 ; T is in Kelvin)
Pure Gases/Mixtures c(O) c(1) C(2)

CH 4 9.27260 x 10- 3 -3.76132x10- 5 4.93066 x 10- 8

N2 7.84980 x 10- 3 - 3.98950 x 10- 5 6.11870x 10- 8
CO 2 2.05130x10- 3 3.48880 x 10- 5 - 8.37030 x 10 - 8
C 2H s - 6.21000 x 10- 2 5.08050 x 10- 4 - 8.85260 x 10- 7
H2 1.04711x10- 3 - 3.64887 x 10 - s 4.67095 x 10- 9
C 3H 8 - 2.70290 x 10- 1 1.71040 x 10- 3 - 2.50010 x 10- s
CO 1.90870 x 10- 3 4.20040 x 10- s -2.56800x10- 8
CH 4 /CH 4 /N 2 8.89649 x 10- 3 - 4.10784 x 10- 5 6.03723 x 10- 8
CH 4/CH 4/C0 2 1.17576x 10- 2 - 4.47615 x 10- 5 5.05481 x10- 8
CH 4/CH 4/C 2H s 1.60923 x 10- 2 - 6.38586 x 10 -5 7.79273 x 10- 8
CH 4/CH 4/H 2 2.33159 x 10- 4 1.02923 x 10- 5 - 2.27208 x 10- 8
CH 4/CH 4/C 3H 8 8.92850 x 10- 3 1.90637 x 10- 5 - 9.79363 x 10- 8
CH 4/CH 4/CO 7.36748 x 10- 3 - 2.76578 x 10- 5 3.43051 x 10- 8
CH 4/CH 4/C 4H 10 2.27707 x 10- 1 -1.39926 x 10- 3 2.21564x 10- s
CH4/CH4/C5H12 3.20344 x 10- 2 -6.91674x10- 5 0.00000
CH4/CH4/CsH14 3.20344 x 10- 2 -6.91674x10- 5 0.00000
CH 4/CH 4/C 7H 1S 3.20344 x 10- 2 - 6.91674 x 10- 5 0.00000
N 2/N 2/CH 4 8.74115xlO- 3 - 4.43880 x 10- 5 6.96584 x 10- 8
N 2/N 2/C0 2 5.52066x10- 3 -1.68609 x 10- 5 1.57169x10- 8
N 2/N 2/C 2H s -1.36584 x 10- 2 1.06972 x 10- 4 -1.79369x10-7
N2/N2/H2 -6.50954x 10- 3 5.19104x 10- 5 - 8.87528 x 10- 8
N 2/N 2/C 3 H 8 6.74630 x 10- 3 -3.65557x10-s - 2.34530 x 10- 8
N 2/N 2/CO 6.15912x 10- 3 -2.72612x 10- 5 3.91226x 10- 8
N 2/N 2/C 4H 10 -1.01202 x 10- 1 7.23716x10- 4 -1.21618x10- s
C0 2/C0 2/CH 4 8.77002 x 10- 3 -1.67108 x 10- 5 -1.51577x 10 -10
C0 2/C0 2/N 2 3.58783 x 10- 3 8.06674 x 10-s - 3.25798 x 10- 8
C0 2/C0 2/C 2H s 1.54623 x 10- 1 -9.65716x10-4 1.56362x10-s
C0 2/C0 2/H 2 8.43475 x 10- 3 - 3.55786 x 10- 5 4.50909 x 10- 8
C0 2/C0 2/C 3 H 8 - 5.94074 x 10- 2 4.30798 x 10- 4 - 6.91157 x 10 -7

C2Hs/C2Hs/CH4 - 8.27905 x 10- 3 1.21913x10- 4 - 2.39958 x 10- 7

C 2H S/C 2H S/N2 - 3.69750 x 10- 2 2.89620x10- 4 -4.94852 x 10- 7
C 2H s/C 2H s/C0 2 1.33240 x 10 -1 - 8.07996 x 10 -4 1.29643 x 10-s
C 2H S/C 2HS/H2 8.43800 x 10- 3 -2.65166x10- 5 3.81345x10- 8
C2Hs/C2Hs/C3H8 - 1.44234 x 10 -1 1.01275 x 10- 3 -1.63277 x lO- s
H 2/H 2/CH 4 -3.08914x10- 3 2.54973 x 10- 5 -4.39625x10- 8
H2/H2/N2 -8.77711 x10- 3 6.39925 x 10- 5 -1.07591 x 10- 7
H 2/H 2/C0 2 8.10002xlO- 3 - 4.84242 x 10 - 5 7.45489 x 10- 8
H2/H 2/C2 H s 2.94870x10- 2 -1.97083 x 10- 4 3.34778 x 10- 7
C3H8/C3H8/CH4 - 8.48395 x 10- 2 6.05618x10- 4 -9.54102x10- 7
C3H8/C3H8/N2 - 8.59670 x 10- 2 5.79776 x 10- 4 -8.77215x 10- 7
C 3H 8/C3 H 8/C02 - 1.49820 x 10 - 1 9.94385 x 10- 4 -1.52037x10- s
C3H8/C3H8/C2Hs - 2.16734 x 10- 1 1.43604 x 10- 3 -2.21182 x 10-s
CO/CO/CH 4 4.36399 x 10- 3 -1.00358 x 10- 5 6.60155 x 10- 9
CO/CO/N 2 4.18924 x 10- 3 -1.26365 x 10- 5 1.36268 x 10- 8
CH 4/N 2/C0 2 9.00338 x 10- 3 - 3.55487 x 10- 5 4.67680 x 10- 8
CH 4/N 2/C 2H s 4.49508 x 10- 3 5.30233 x 10 -7 - 1.52576 x 10 -8
CH 4/N 2/H 2 3.60589 x 10- 3 -1.45437 x 10- 5 2.04722 x 10- 8
CH 4/N 2/C 3 H 8 -1.87767x10-2 1.42145x10- 4 -2.21855x 10- 7
CH 4/C0 2/C 2 H s -1.76574x10-3 5.97684 x 10- 5 - 1.23785 x 10 - 7
CH 4/C0 2 /C 3 H 8 - 6.23055 x 10- 2 4.59485 x 10- 4 -7.33765 x 10- 7
CH 4/C 2H s/H 2 -1.16095 x 10- 3 2.63812 x 10- 5 -5.12600x10-8
CH4/C2Hs/C3He - 1.89480 x 10 -1 1.39040 x 10 - 3 -2.38860x10-S
CH4/C2Hs/C4HlO - 1 .89480 x 10 -1 1.39040 x 10- 3 -2.38860x10-s
CH 4/C 2H s/C 5H 12 - 1.89480 x 10 - 1 1.39040 x 10- 3 -2.38860x10- s
N 2/C0 2/C 2H s -7.68174x10-s 3.85226 x 10- 5 - 8.17506 x 10- 8
N 2/C0 2/H 2 2.69108 x 10- 3 - 5.96904 x 10- s 5.40142 x 10- 9
N 2/C02/C3 He -3.01719x 10- 2 2.22387 x 10- 4 -3.50693 x 10- 7
N2/C 2H S/H2 - 2.97743 x 10- 4 1.68641 x10- 5 - 3.30680 x 10- 8
N2/C 2H s/C 3H a - 5.06529 x 10- 2 3.63400 x 10- 4 -5.77241 x 10- 7
C0 2/C 2H s/H2 -5.16317x10- 3 5.79276 x 10- 5 -1.05347x10- 7
C0 2/C 2H s/C 3H e - 9.44235 x 10- 2 6.71721 x 10- 4 -1.08080 x 10-s

346 SPE Production Engineering, August 1991

 TABLE 5-S0URCE OF DATA SETS· TABLE 6-COMPARISON OF PERFORMANCES OF
GERG VIRIAL AND SUPERCOMPRESSIBILITY EQUATIONS
Data Sets
Synthetic gases Natural gases Pressure Number RMS Error (%)
British Gas 1 (7) 3 (64) Range of GERG Supercompressibility
Gasunie 1 (22) 28 (1,386) (MPa) Points Equation Equation
Gaz de France 12 (72) 5 (50) o to 8 4,043 0.054 0.083
Ruhrgas AG 6 (391) 35 (1,989) 8 to 12 443 0.065 0.223
Literature 0(0) 13 (997) o to 12 4,486 0.055 0.105
Total 20 (492) 84 (4,486)
"Number of data sets given in parentheses.
200~-----------------------.

A global comparison of the predicted data with all experimental all expo data with X(02~ 0.04
( 477)
compressibility factors for natural gases in the GERG data bank,
a comparison for the data grouped into temperature and pressure
....c:
VI

QI
domains, and a comparison for natural gases grouped according e
to composition were done to study the accuracy of the GERG viri- ....QI
al equation. The results presented here use the following definitions:
0-
X

-
flz=(ZC-ZE)/ZE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (11) QI

for relative deviation (also called Z error) and 100

0
....
£=(~ i; flzp)I/2 .............................. (12) QI
.a
N i=1 e
:::l
c:
for RMS error.

Global Comparison of GERG Virial Equation. The average RMS

error for the difference between the compressibility factors pre-
dicted by the GERC virial equation and all measured gas compress-
o .l..lL....lL,....lL....lL....l!.......!L/"Z

ibility factors in the data bank is as follows.

1. £=0.055% for all pressures between 0 and 12 MPa. It is, how-
ever, pressure-dependent (see Table 6).
2. £=0.054% for pressures between 0 and 8 MPa.
3. £=0.065% for pressures between 8 and 12 MPa.
These results apply to all measurements. A more detailed analy-
sis showed that the RMS error is only 0.023% between 0 and 8
MPa and 0.059% between 8 and 12 MPa if the comparison is limited
to data obtained with the Burnett apparatus and the optical method 9
(Table 3). The distribution curve of the percentage of relative devi-
ation between predicted and measured compressibility factors for
all experimental data points shows the normal shape of an error
distribution curve, even though compressibility factor deviation of
the GERG virial equation tends to be slightly negative (see Fig. 1).
The GERG virial equation predicts compressibility factors ex-
cellently, although predicted data differ in some 250 cases by more
than ±0.1 % from the measured value (the total number of ex-
perimental data points is greater than 4,400). Most of the experimen-
tal data with deviations greater than ±0.1 %, however, were
measured with the DEH Z meter or were taken from the literature.
These data had experimental errors greater than those of the data
measured with the Burnett or optical method. 9 Even in these cases,
however, the deviation remains, with very rare exceptions, within
±0.2 %. Thus, within the combined uncertainties of the experimen-
tal data and the target accuracy of the equation, the experimental
and predicted data agree very well.
Analysis by Pressure/Temperature Domains. Analysis of the
GERG virial equation can be refined further by comparison of the
results in certain pressure and temperature domains. The agree-
ment between the predicted Z values and all the experimental data
for the natural gases grouped into such pressure and temperature
domains will be directly influenced by the experimental uncertain-
ties of the available Z data and the accuracy of the equations.
For the analysis, the experimental data were grouped in several
pressure and temperature domains, and the RMS errors for all the
experimental data points within these domains were plotted (see
Fig. 2). The RMS error remains below 0.08% as long as the pres-
sure is below 6 MPa or above 10 MPa. The RMS error exceeds
this value only in the pressure range of 6 to 10 MPa. A maximum
error of 0.10% occurs in the 265 to 275-K temperature interval.
-0.2 -0.1 o 0.1 0.2
percentage relative deviation
Fig. 5-Relative error distribution of the experimental z data
for natural gases containing more than 4% CO 2'
Analysis of the data measured with the optical or Burnett apparatus
shows only an RMS error of less than 0.07 % over the whole pres-
sure and temperature range (Fig. 3). This indicates the high con-
sistency and accuracy of these data measured up to 12 MPa. It also
shows that up to 8 MPa, the RMS error is less than 0.05 %, so twice
the RMS error is still less than O. 1%, the target accuracy . For higher
pressures, the RMS error is still less than 0.07%.
So far, RMS errors averaged over all natural gases have been
considered. It might be, however, that for certain natural gas com-
positions, the predicted Z values are less accurate. To test the sen-
sitivity of the GERG vi rial equation to gas composition, the results
are compared for particular natural gas data sets.
Accuracy for Groups of Natural Gases. The distribution of the
relative deviation of predicted values of Z compared with the meas-
ured values is presented in Figs. 4 through 7 for four groups of
natural gases. The measured data cover the whole range of applica-
bility of the GERG virial equation, as shown in Table I.
Natural Gases Containing More Than 9.5% Nitrogen. The dis-
tribution curve is highly asymmetrical for this group (Fig. 4). The
relative deviation tends to be negative, with a maximum between
- 0.075 and - 0.05 %. This maximum corresponds to the Gasunie
data, all of which have negative deviations. A second maximum
exists between 0 and 0.025 %, which corresponds to very recent
and highly accurate data covering the entire pressure range of 0
to 12 MPa for temperatures between 270 and 330 K. The mean
deviation is slightly positive and the RMS error has a value of only
0.03 % for these data. Therefore, compressibility factors of natu-
ral gases with high nitrogen content are predicted well by the GERG
virial equation.
Natural Gases Containing More Than 4% CO 2 , For these
gases, the distribution is slightly displaced toward more positive
values (Fig. 5) because of a single gas with a mean deviation of
+0.075%. The other gases have mean deviations of less than
0.025%.

SPE Production Engineering, August 1991 347

 400 400
all expo data all expo data wi th XH, .. 0,02
wi th XC,H," n08 (690 )
....c:
."
(1277)
(4 outside! 0.25'%)
....c:
."
14 outside! 0.25%)

QJ QJ
E E
'- '-
QJ QJ
0.. 0..
)(

-
)(

-QJ QJ

200 200
0 0

'- '-
QJ QJ
.0 .0
E E
:::J :::J
c: c:

o -0.2 -0.1 o 0.1 0.2

percentage relative deviation percentage relative deviation

Fig. 6-Relative error distribution of the experimental z data Fig. 7-Relative error distribution of the experimental z data
for natural gases containing more than 8% ethane. for natural gases containing more than 2% hydrogen.

Natural Gases Containing More Than 8% Ethane. The distri-
bution curve for this group is fairly symmetrical (Fig. 6). A more
detailed analysis shows that inconsistencies appear for certain ex-
perimental data in this group. However, the most recent compress-
ibility factor measurements are extremely well-predicted by the
equation at 270 to 330 K and 0 to 12 MPa.
Natural Gases Containing More Than 2% Hydrogen. These
gases include mixtures of natural and coke-oven gases. For these
gases, the distribution curve is slightly displaced toward negative
values (Fig. 7). The number of points for which the relative devia-
tion is greater than 0.1 % is less than 6% of the total. Therefore,
compressibility factors of natural gases containing hydrogen are
correctly modeled by the OERO virial equation.
Comparison With Other Methods
The influence of pressure on deviation between compressibility fac-
tors predicted by the OERO virial equation and measured values
for several gases along the 270-K isotherm is small (Fig. 8). Even
for pressures between 8 and 12 MPa, the differences are less than
0.1 % for these natural gases (for the natural-gas/coke-oven-gas mix-
ture below a pressure of 11 MPa). This excellent agreement is satis-
factory because the differences between predicted and measured
compressibility factors are invariably highest for low temperatures
and high pressures, while they are relatively small for high tem-
peratures. For the modified version of the AOA NX-19 method, \,2
the deviations for the same experimental data may be as great as 1 %.

,..
0.4 ()
T=270K
0.4 fYt,()
() natural gas
1=270 K
0.3 ()
coke oven gas ()
0.3 ()

0~
f) Drohne gas
e NAM gas
0.2 ('
'-
. 0.2 o Soviet gas 0.1
()()
fYt,'
\.II'
0~
0
.. TENP gas
4~iiii_~~~e~~eeee~
'-
'-
QJ 0.1 • Ekofisk ~
0.0
0
N '-
'- I I!s~ ~f) OOOO~~
natural gas/ •• S f)f)f)f)f)
QJ
0.0 N -0.1 I-()
coke-oven gas • ....
f) Drohne gas . . ...',... ,r.
-0.1 -0.2 I-
e NAM gas 4'--
-0.2
o Soviet gas
-0.3 - .. TENP gas •• •
-0.3
• Ekofisk gas
C
•••••
0 4 8 12 -04. 0 4 8 12
pressure p, MPa pressure p, MPa

Fig. 9-z errors, difference between calculated z values (GRI

Fig. 8-z errors, difference between calculated z values (GERG supercompressibility equation) and experimental z values
virial equation) and experimental z values (measured with the (measured with the grating interferomet~r) for six natural
grating interferometer) for six natural gases at 270 K. gases at 270 K.

348 SPE Production Engineering, August 1991

 Reasonable results are also obtained by the recently developed
supercompressibility equation developed by Starling 14 under the
sponsorship of the Gas Research Inst. (GRI) in cooperation with
the AGA's Transmission Measurement Committee.
For four groups of natural gases, the deviations are below ± 0.1 %.
(Fig. 9). For Ekofisk gas, however, the deviation is -0.2% at 8
MPa and 270 K, and -0.35% at "" 10 MPa. For the natural-
gas/coke-oven-gas mixture, the deviations are at least 0.4 %.
Thus, the supercompressibility equation seems less accurate than
the GERG virial equation, especially for high pressures. This re-
sult is confirmed by calculation of the RMS errors of the differ-
ence between the calculated and measured z values for all the natural
gas data in the GERG data bank (see Table 6).
Conclusions
The GERG equation predicts compressibility factors within ±O.l %
for natural gases for pressures up to 12 MPa (and for natural-
gas/coke-oven-gas mixtures for pressures up to 10 MPa) for tem-
peratures between 265 and 335 K. The accuracy of the equation
is not yet fully established for all possible gas compositions. The
GERG virial equation is, of course, valid only if the natural gas
is in the truly gaseous phase.
A simplified version, for which the input data are four charac-
teristic properties chosen from relative density, gross calorific value,
and the molar fraction of CO 2 , nitrogen, and hydrogen, has been
derived from the GERG virial equation based on a full gas analy-
sis. The structure and performance of this simplified GERG virial
equation is discussed in a companion paper. 15
Nomenclature
MO) = zero-order term in expansion of B in temperature
(Eq. 5), m 3 /kmol
b(l) = coefficient of first-order term in expansion of B in
temperature, m 3 Ikmol' K
b(2) = coefficient of second-order term in expansion of B
in temperature, m 3 Ikmol . K 2
B = second virial coefficient, m 3 Ikmol
c(O) = zero-order term in expansion of C in temperature
(Eq. 6), m 6 /kmoJ2
c(l) = coefficient of first-order term in expansion of C in
temperature, m 6 Ikmol 2 • K
c(2) = coefficient of second-order term in expansion of C
in temperature, m 6 IkmoJ2 . K 2
C = third virial coefficient, m 6 /kmol 2
E = fractional (or percentage) RMS deviation of data set
n = number of components in mixture
N = number of data points in set
p = (absolute) pressure, Pa
R = universal gas constant (8314.51), J Ikmol . K
T = absolute temperature, K
x = mole fraction
z = compressibility (or compression) factor
.iz = relative deviation of a single data point (Eq. 11)
~ = coefficient in combining rules for virial coefficients
(Eq. 10)
Pm = molar density, kmollm 3
Subscripts
C = calculated
E = experimental
= ith component in a mixture or ith data point in a set
ij = binary interaction
ij k = ternary interaction
j = j th component in a mixture
SPE Production Engineering, August 1991
k = kth component in a mixture
M = mixture
Acknowledgments
The research leading to the results discussed here was sponsored
by the GERG. The GERG virial equation based on the coefficients
of the temperature expansion for the second and third virial coeffi-
cient is available on an IBM-PC-compatible diskette from Ruhr-
gas, Gaz de France, Distrigaz, British Gas, Gasunie, SNAM, and
the van der Waals Laboratory.
References
1. Heming, F. and Wolowski, E.: "Kompressibilitiitszahlen und Real-
gasfaktoren von Erdgasen nach neuen amerikanischen Berechnungs-
methoden," Ges. Ber. Ruhrgas (1966) 15.
2. Jaeschke, M. and Harbrink, B.: "Korrektur des AGA NX-19-mod-
Rechenverfahrens fiir Realgasfaktoren von Erdgas," H. gwf-GasIErd-
gas (1982) 123, No. I, 20-27.
3. GERG Information Committee on the Thermodynamic Properties of
Natural Gas: "Accurate Measurement and Prediction of the Compress-
ibility Factor of Natural Gas Mixtures for Use in Transmission," paper
IGu/C4-82 presented at the 15th World Gas Conference, Lausanne
(1982).
4. Melvin, A. et al.: "Predictive Methods for the Compressibility Fac-
tors of Natural Gases in Transmission," Proc. of 1986 IGRC, Toron-
to (1987) 438-48.
5. Michels, J.P.J., Schouten, J.A., and Jaeschke, M.: "The Determina-
tion of the Second and Third Virial Coefficient From PVTx-Data of
Binary Systems," Inti. 1. Thermophysics (1988) 9, No.6, 985-92.
6. Schouten, J.A., Michels, J.P.J., and tenSeldam, C.A.: "Updating of
the GERG Equation Based on Full Gas Analysis," report 8709-1, van
der Waals Laboratorium, U. of Amsterdam, The Netherlands (1987).
7. McGregor, D.R. et al.: "An Interaction Model for Third Virial Coeffi-
cients Which Provides Simplified Equations for Mixtures," Fluid Phase
Equilibria (1987) 35, 153-64.
8. Burnett, E.S.: "Compressibility Determinations Without Volume Meas-
urements," ASME Trans., 1. Appl. Mech. (1936) 58, A 136-40.
9. Jaeschke, M., Hinze, H.M., and Scheuren, H.: "Measurement of Com-
pressibility Factors of Natural Gases up to High Pressure (30 MPa),"
Report 6500.024, Ruhrgas AG Applied Physics Section, Dorsten, Ger-
many (May 1987).
10. Achtermann, H.J. et al.: "Direct Determination ofthe Second Refrac-
tivity Virial Coefficient of Methane, Nitrogen and Five of Their Mix-
tures," Inti. 1. Thermophysics (1986) 7, No.2, 357-66.
11. Bose, T.K. et al.: "Improved Method for the Precise Determination
of the Compressibility Factor from Refractive Index Measurements,"
Rev. Sci. Instrument (1986) 57, No. I, 26-32.
12. Jaeschke, M. et al.: High Accuracy Compressibility Factor Calcula-
tionfor Natural Gases and Similar Mixtures by Use ofa Truncated Virial
Equation, GERG Technical Monograph TM 2 (1988) and Fortschritt-
Berichte VDI, Series 6, No. 231 (1989).
13. Jaeschke, M. and Humphreys, A.E.: The GERGDatabank ofHighAc-
curacy Compressibility Factor Measurements, GERG Technical Mono-
graph TM 4 (1990) and Fortschritt-Berichte VDI. Series 6, No. 251,
(1991).
14. Starling, K.E.: "Compressibility and Supercompressibility for Natu-
ral Gas and Other Hydrocarbon Gases," AGA Transmission Measure-
ment Cominittee Report No.8, America Gas Assn. (Dec. 1985).
15. Jaeschke, M. etal.: "Simplified GERG Virial Equation for Field Use,"
SPEPE (Aug. 1991) 350-55.
Author photographs and biographies appear on Page 355.
SI Metric Conversion Factors
OF (OF+459.67)/1.8 K
psi x 6.894 757 E+OO kPa
SPEPE
Original SPE manuscript received for review June t3, t988. Paper accepted for publica-
tion May 22, t990. Revised manuscript received April 9, t990. Paper (SPE 17766) first
presented at the 1988 SPE Gas Technology Symposium held in Dallas, June 13-15.
349
Discussion of Accurate Prediction of Compressibility
Factors by the GERG Virial Equation and of Simplified
GERG Virial Equation for Field Use
J.L. Savidge, SPE, Gas Research Inst.
As with most equation-of-state (EOS) modeling work, improvements
come iteratively on the basis of the availability of improved funda-
mental data. The work of Iaeschke et at. is widely commended for
the significant contribution made to improving both the fundamental
data base on natural gases and to providing the gas industry with
exceptionally high-quality models representing those systems. This
was achieved through a meticulous data evaluation/modeling effort
and a comprehensive experimental program cooperatively sponsored
by the research consortium GERG.
The two papers on high-accuracy virial models for natural gases
were written shortly after new, proprietary natural gas PVT data
became available for development of the GERG virial equations.
The AGA Report No.8 equation (1985 publication) the authors
reference did not have the benefit of these data for its development.
Iaeschke et at. make the point of noting the poor performance of
the AGA Report No.8 model on coke-oven gases. This is not sur-
prising because those systems were not included in the original AGA
Report No.8 modeling work. After the writing of these papers,
the AGA Report No.8 method was revised with the new PVT data
provided by Iaeschke et at.
The summary statistics on the revised AGA Report No.8 method
and the GERG vi rial equations method indicate that the methods
produce virtually identical results in the low-pressure regime report-
ed in Iaeschke et at. 's papers. Overall statistical comparison ofthe
methods for 4,487 data points gives an average absolute deviation
of 0.035% for revised AGA Report No.8 and 0.038% for the
GERG equation, standard deviations and RMS deviations around
0.05%, and maximum deviations for both methods near -0.4%.
The virial model has advantages to many users because of its sim-
356
plicity, but it does have limitations, identified by the authors. The
application guidelines they provide should be strictly followed. This
is important because of the rapid deterioration of accuracy as the
truncated virial model is extrapolated beyond the specified pressure
or temperature limits. This is an inherent problem with a truncated
third virial coefficient model. It points out a major difference be-
tween the AGA Report No.8 model and the GERG model: the form-
er is designed to cover a significantly broader range of gas-phase
conditions than those occurring in natural gas transmission lines.
To accomplish this, the AGA Report No.8 model uses higher-order
density terms beyond the third virial coefficient terms. A hybrid
mixing model is applied to avoid inherent problems with the mix-
ing of high-order virial terms. The AGA Report No.8 approach
is more applicable to higher pressures and densities. This will be
a concern to engineers working with high-pressure gas reservoirs,
gas production, gas processing, high-pressure metering, or high-
pressure end-use applications. For interested readers, the revised
AGA EOS is currently available as a Gas Research Inst. Annual
Report (Report No. GRI-91/0l84, "GRI High Accuracy Natural
Gas Equation of State for Gas Measurement Applications"). It will
be published shortly as a revised AGA Report No. 8 on Natural
Gas Compressibility.
The simplified GERG virial equation and the revised AGA Report
No.8 equation will constitute a new ISO standard and AGA stan-
dard to compute natural gas compressibility factors for high-
accuracy gas-metering applications. Both standards are in prepa-
ration now and will be available soon.
(SPE 23568) SPEPE
SPE Production Engineering, August 1991