Abstract
Reservoir fluids often show complex compositional behaviors in single columns in equilibrium due to combinations of gravity, capillary and chemical forces. Frequently non equilibrium or non stationary state conditions are also encountered, for instance due to thermal forces acting. Recognizing these behaviors downhole is a complex process that requires a greater number of data points, fluid samples and associated laboratory analysis.
Pressure gradients with wireline formation testers are traditionally used to evaluate fluid density, fluid contacts, and layer connectivity in exploration settings. This information is today supplemented by downhole fluid analysis (DFA) measurements to reveal possible reservoir fluid heterogeneities.
Although these fluid complexities have been largely recognized, conventional pressure-depth plot and pressure gradient analysis are still performed with traditional straight line regression schemes. This process may however be misleading as fluid compositional changes and compartmentalization give distortions in the pressure gradients, which lead to erroneous interpretations of fluid contacts or pressure seals. Hence the models imposed on the pressure data to calculate pressure gradients need to incorporate a rigorous mathematical approach to respect all data available, so as to follow an objective assessment of reserves and reservoir architectures.
This paper presents a method to use combined repeated pressure and in-situ fluid measurements to provide a simple model of vertical fluid distributions, looks at the different regression schemes that can be imposed on pressure data to calculate fluid gradients with their associated uncertainties and concludes on an optimal fit approach. This data integration then allows making assessments and quality control of the different measurements and conclusions about the relevant reservoir heterogeneities.
The method is illustrated with a published case study [1] from a North Sea appraisal well, where a large compositional gradient has been observed with in-situ fluid measurements. An equation of state is elaborated from a sample and its PVT experimental results, and a compositional gradient is parametized using the DFA observations at the different depths. A polynomial fit is then given to the distributed pressure measurements and the obtained fluid density variations are compared to the fluid model ones.