Rock Type Modelling of a Heterogeneous Tidal Reservoir of the Tilje Formation

Heterolithic tidal reservoirs, like those of the Tilje Formation, are challenging to model. In addition to primary facies, diagenetic overprint adds to the heterogeneous character of the reservoir. A rock type modelling approach is presented to enable the use and integration of conventional core data, whole core computer tomography (CT) scans, wireline data and well test data to characterise the reservoir. Whole core CT data oriented to image log has allowed paleocurrent data, depositional facies and rock types to be interpreted at very fine scale. The combination of these data allows a better understanding of permeability distribution and anisotropy. Permeability for given porosity class can have several orders of magnitude difference. The best rock type has permeability in excess of 1000 mD due to the presence of chlorite coating around grains which prevents the development of quartz overgrowths expected at the burial depths of the reservoir (4.5-5 km TVDSS). Rock type with reduced grainsize and/or thicker chlorite coating have reduced pore throats reducing permeability to 10 to 100 mD. In the absence of chlorite coating permeabilities are below 1 mD. Well test interpretation suggests the presence of barriers close to the well bore with good connectivity in alternate directions. It is not known if the barriers are due to sealing faults or due to depositional or diagenetic features. The geological model has been built to ensure that the heterogeneity observed in logs and core and the response from well test is captured to allow production and injection well scenarios to be tested. Rock types have been classified on the basis of Rock Quality Index (RQI) which is related to both original depositional textures (sorting and grain size) and diagenesis. Once defined at the plug scale whole core CT scans have been used to propagate the rock type in cored intervals. The resulting rock type curve was then used as a training data set for machine learning algorithms to populate rock types in un-cored intervals. Permeability thickness (kh) from well tests has been used to quality control and calibrate the rock type model. Once propagated to wells a multi scenario 3D rock type model was built.