Experimental study of multiphase fluid flow and trapping in porous carbonate rocks

Multiphase immiscible fluid flow in porous media is an important physical phenomenon that impacting many natural and industrial processes such as groundwater migration and enhanced oil recovery. The scope of this study is pore-scale flow processes because it is the pore-scale behaviour of immiscible fluids in a porous material that controls the macro scale behaviour. The porous media focused on this study are carbonate rocks because they are of global importance as they contain about 50% of the world’s hydrocarbon reserves and are also a major host to groundwater resources globally. Carbonate rocks are much less understood than sandstones because they often have much more heterogeneous properties in porosity, pore size distribution and pore network structure. The experimental approach of this research consists of a series of core-flooding experiments using both in-house and synchrotron-based X-ray microtomography (X-ray µCT) facilities. The in-house X-ray micro CT imaging is used to image static rock samples. The synchrotron µCT with its high photon flux enables active processes of fluid flow to be imaged in-situ and in real-time, so this 4D time-resolved imaging technique was used to observe the impact of porous media properties on actual flow events in the experiments. Displacement processes that occur in a mixed-wet Indiana limestone plug under brine injection, oil injection and steady-state injection were investigated and imaged with synchrotron µCT. Many of the previous works have done unsteady-state injection. In this study, direct observation of two-phase flow in carbonate rocks and measurements of saturation and fluid topological connectivity are presented in both steady-state and unsteady-state flow. Fluid displacement and trapping processes, as well as the fluid connectivity, were compared during the three injection stages. The result confirmed connectivity hysteresis during drainage and imbibition. In the other core-flooding experiment, a carbonate ’shrub’ sample from the Brazilian Pre-salt reservoir was used, together with dead crude oil from the Pre-salt Lula field. The displacement processes were imaged with in-house conventional X-ray µCT. In the Pre-salt ’shrub’ sample, fluid displacement using crude oil and mineral oil was compared. The crude oil achieved higher saturation than the mineral oil during drainage and remained less residual saturation after the imbibition of brine. The present work demonstrates that the pore-scale fluid displacement and trapping in carbonates are much more complex than that occur in idealised or relatively homogeneous media such as sandstones. On the other hand, pore-scale multiphase flow phenomenon was numerically modelled, and validated by our experimental observation. The model simulation was studied by the collaborating research group at Heriot-Watt University by Dr Julien Maes. In this research, a Roof snap-off event was identified from the experiments, and the event was modelled and repeated using a direct numerical simulation method. A deep learning method was implemented to improve the image segmentation workflow. The deep learning approach increased the temporal efficiency of the conventional segmentation workflow by ten-fold with equally good accuracy and is proved to have a strong resistance to ring artefacts. It also minimised human intervention during segmentation and allow unsupervised processing. The deep learning approach provides a promising tool for the image processing of extra-large µCT datasets. To conclude, the study increased the understanding of immiscible multiphase flow in porous carbonates, and provided a promising approach of processing and segmenting large-scale X-ray µCT datasets.