Mudlogging has turned mud gas measurements (or Gas While Drilling, GWD) to a sufficient level of reliability to consider the C1 to C5 (C6) surface gas compositions as a reliable picture of the hydrocarbon fluid content of the mud. Provided all drilling artifacts are properly addressed, it comes to a real continuous fluid composition log along the well. Such a log is however limited to the light end of the fluid and, consequently, cannot provide straightforward conclusions on the fluid nature and properties. In the meantime, Downhole Fluid Analysis (DFA) provides real time measurements of fluid properties while pumping out the reservoir fluid at selected stations, such as a quantitative lumped composition (i.e. C1, C2-C5 and C6+ for instance), GOR and live downhole fluid density. When fully representative, these measurements can help identifying the fluid nature with the restriction of the vertical representativity of such stations.
The development presented in this paper consists in associating locally the decontaminated DFA measurements and the corresponding stacked quantitative mud gas data to fit a chemical and physical fluid model. This latter is a simplified average molecular fraction pattern law that statistically describes the abundance of each cut based on compositional invariants, expresses the overall fluid column equilibrium and associates volumetric properties to each cut. All the unknown parameters can be auto-adjusted by inversion provided a full C1-C5 molecular composition, C1-(C2)-C3/5-C6+ weight percents, GOR and downhole density are provided by the different techniques. Then, the model can predict the other fluid properties (such as molecular weight, FVF, API…) and propagate them along the well section thanks to the quantitative mud gas data alone.
The quantitative correlation technique was applied to various reservoirs and fluid contexts. A thin low GOR mono-fluid case tested the model robustness at predicting non calibrated properties. A thick hydrocarbon column was used to evaluate the model sensitivity to segregation. And finally, a multi-layered system is presented to validate the prediction reliability of the technique in a compartmentalized context.
Deriving a fluid properties log along a section from quantitative mud gas is an attractive option as such data are systematically acquired on all wells at reasonable cost. But a real fluid prediction becomes reliable only when the input data are fully consistent and quantitative, and provided the fluid is simple enough to comply with the molecular fraction pattern model (it potentially excludes highly biodegraded, multi-sourced or reworked fluids…). Then, this approach can have significant impact at exploration or appraisal stage to refine sampling programs and assess the quantitative impact of a sample in the fluid column understanding, qualify precisely reservoir continuity even in difficult acquisition contexts (low permeability formations, clayey reservoirs, thin sands, multilayered reservoirs…) and early estimate average fluid properties, heterogeneities and gradients.