This paper presents preliminary laboratory measurements of the natural gas compressibility or z-factor at high-pressure, high-temperature (HP/HT) reservoir conditions. We conducted laboratory studies to measure gas density and z-factor for several dry gas mixtures at pressures up to 20,000 psia and temperatures of 300°F and 400°F. For our study, we define a dry gas as one that remains in the single hydrocarbon (gas) phase during the entire isothermal pressure depletion path from the reservoir through surface conditions. We also measured the effects of CO2 in the gas phase on gas density and z-factor at HP/HT conditions. Several gas mixtures contained CO2 concentrations up to 20 mol% each.
We then compared the new laboratory data with two equation-of-state (EOS) models used by the petroleum industry to compute z-factor. In combination with our evaluation of the EOS models, we also assessed several mixing rules and empirical correlations for estimating pseudocritical properties of gas mixtures. Results of our comparisons indicate the following:
For gases with no CO2, the Hall & Yarborough EOS  combined with the Sutton [2005, 2007] pseudocritical property correlations were the most accurate, especially in the higher pressure range and for both temperatures evaluated. The Dranchuk & Abou-Kassem EOS  was the least accurate regardless of the mixing rule or empirical correlation used to estimate pseudocritical properties;
Although the Hall & Yarborough EOS combined with the Sutton pseudocritical property correlations were the most accurate for gases with no CO2 at 400°F, the errors were still more than 300% higher than those for the same gas mixtures at 300°F. These differences suggest neither EOS has been properly "tuned" to data at higher temperatures and validates our concerns about extending the Standing-Katz correlations numerically to higher pressures and temperatures;
When combined with the Hall & Yarborough EOS, both the Kay  and the Stewart-Burkhardt-Voo  mixing rules were surprisingly quite often the second most accurate models for estimating z-factors at HP/HT conditions; and
For gas mixtures with CO2, the Hall & Yarborough EOS was again the most accurate when combined with the Wichert & Aziz  correlations for CO2 and the Sutton pseudocritical property correlations. Conversely, the Dranchuk & Abou-Kassem EOS was the least accurate for all mixing rules or empirical correlations.