#19 Reservoir Fluid Characterization

Calsep will over the next couple of months post a series of TechTalks about simulation of fluid properties covering: Basic fluid characterization, Mud contaminated samples, Common EoS, Fluids with water, and CO₂ rich fluids.

This first TechTalk is about characterization of a single clean reservoir fluid composition.

Fluid characterization was a hot topic in the early nineteen eighties, and no wonder why. Most reservoir fluid compositions stopped at C_7+. The molar distribution of the C₇₊ components was pure guesswork, and it was unclear what critical properties (T_c , P_c, and acentric factor) to assign to the C₇₊ carbon number fractions.

Keith Coats was one of the pioneers in reservoir fluid characterization. In an article from 1985 [1] he outlined the challenges encountered in the early 1980s to just characterize a single gas condensate fluid composition. Coats struggled not only with a compositional analysis ending at C₇₊, but also with slow and numerically unstable flash and phase envelope algorithms, and he used an equation of state that was unable to provide accurate liquid densities.

Today the numerical challenges have been overcome and the techniques for analyzing reservoir fluid compositions have improved. It is now industrial standard to report reservoir fluid compositions to C₃₆₊. The problem with poor liquid density predictions was solved when Peneloux and coworkers in 1982 suggested to extend the SRK and PR equations with a volume shift parameter.

When compositions to C₂₀₊ became available in the mid nineteen eighties [2], they revealed a linear relation between ln(mol%) and C₇₊ carbon number. This observed exponential decay might initially have appeared to be a coincidence, but later turned out to be founded on chemical reaction theory [3]. Combining the relation between ln(mol%) and carbon number with today’s extended compositional analyses to C₃₆₊ has made it is straightforward to split the plus fraction into carbon number fractions. The trend line for C₇-C₃₅ is just to be continued as sketched in Figure 1.

For biodegraded fluids, the fractions from C₇ to ~ C₁₅ do not follow the linear trend in Figure 1, as the paraffins in that range have been consumed by bacteria, but the linear trend is seen again from C₁₆ onwards and the C₁₆-C₃₅ molar distribution is sufficient to establish the linear relation needed to split the C₃₆₊ fraction into carbon number fractions.

T_c, P_c, and ω are well-known for components till around C₁₀, but must be estimated for heavier fractions. The trend seen for C₇ – C₁₀ shows that T_c increases and P_c decreases with molecular weight as sketched by the red lines in Figure 2. With T_c and P_c known till around C₁₀, the starting points for the T_c and P_c relations versus C₇₊ carbon number are established. What remains is to determine the right slopes of the T_c and P_c  curves. The T_c and P_c correlations in PVTsim Nova will provide the right qualitative curvature, and PVTsim’s plus fluid regression will as outlined with black lines in Figure 2 fine tune the slopes by rotating the T_c and P_c curves around the values of C₇ to get the slopes that provide the best match of the measured PVT data.

The acentric factor (ω) is a bit more complex, but also less influential on the fluid properties than T_c and P_c. For n-paraffins ω increases with molecular weight, but the concentration of n-paraffins decreases with molecular weight. The C₅₀₊ fraction essentially consists of iso-paraffins, naphthenes and aromatics, which components have lower acentric factors than n-paraffins. The acentric factors of the C₇₊ carbon number fractions are therefore seen to increase with molecular weight till ~C₅₀, while a slight decrease is seen for heavier components.

The last parameter is the binary interaction parameter (k_ij). A k_ij > 0 will reduce the simulated attraction between molecules of types i and j. For hydrocarbon pairs, it is recommended to use k_ij = 0 as this will maintain the right balance between the attractive and the repulsive terms of the equation of state also at temperatures other than the one for which experimental PVT data has been tuned to. A slightly positive k_i is needed for component pairs counting a non-hydrocarbon (N₂, CO₂ and H₂S). For a CO₂-hydrocarbon pair a k_ij of 0.08-0.12 is appropriate.

Finally, the C₇₊ fraction must be split into a manageable number of pseudo-components, each of approximately equal weight amount, and the characterization is complete.

In PVTsim Nova the characterization and regression to PVT data are automated in a so-called Auto EoS option. If the user further takes advantage of the option for automated import of PVT reports in PRODML XML format (see Figure 3), development of an EoS model can be done in a few minutes.

Figure 3 PVTsim Nova option for automated import of PVT reports in PRODML XML format.

As can be seen from Figure 4, the Auto EoS option in PVTsim only requires input of the number of components in the final EoS model and the importance  to be assigned to each data type (High, Medium, Low, None).

A few seconds later the EoS model will be made and is ready for use in simulations and for export to external software products (Eclipse, OLGA, Hysys, etc.). Plots are shown of the match of experimental PVT data as exemplified in Figure 5.

As can be seen from the above, it is relatively straightforward to characterize a single reservoir fluid composition, but fluid characterization can still present challenges. Many samples are mud contaminated, which means that the measured PVT data is not representative of the pure reservoir fluid. Many PVTsim users demand field-wide EoS models and models that can handle water and fluids with a high CO₂ content. These challenges will be addressed in TechTalks in the coming months.

References

1. Coats, K. “Simulation of gas condensate reservoir performance”, Journal of Petroleum Technology, October 1985, pp 1870-1886.
2. Pedersen, K.S., Thomassen, P., Fredenslund, Aa., “Thermodynamics of Petroleum Mixtures Containing Heavy Hydrocarbons. 1. Phase Envelope Calculations by Use of the Soave-Redlich-Kwong Equation of State”, Ind. Eng. Chem. Process Des. Dev. 23, 1984, pp. 163-170.
3. Sørensen, H., Pedersen, K.S., Christensen, P.L., “Method for Generating Shale Gas Fluid Composition from Depleted Sample”, Forth International Acid Gas Injection Symposium (AGIS IV), Calgary, September 24–27, 2013.