Integrated Study of Carbonate Reservoir Characterization in Balkassar Field, Potwar Plateau, Pakistan

Abstract

The carbonate reservoir characterization is very challenging due to its diverse nature. For economical production from a fractured reservoir, a characteristic analysis of the fracture parameters (i.e. density and orientation) is essential to improve the fluid flow during extraction. Furthermore, the complexity in their internal structure, depositional settings, and diagenetic history deems the application of conventional methods for reservoir evaluation, inappropriate. The identification of conductive fractures prior to drilling is another significant factor in characterizing carbonate reservoirs. A set of advanced techniques must therefore be used which deals with delineating the fracture network system (orientation, distribution, and extension of fractures) within the reservoir. This dissertation proposes a proper rock physics model and utilizes an advanced machine learning approach with seismic attributes analysis to characterize a carbonate reservoir. The crest of hydrocarbon producing fault-bounded Balkassar anticline in Northern Potwar, Upper Indus Basin, Pakistan has been selected as the case study representing a potential zone for development of fractures at reservoir level (Sakesar Limestone). The methodology consists of the interpretation of 3D post-stack seismic data acquired in the Balkassar area and conventional wireline log data to demarcate the reservoir containing fractures. The petrophysical analysis was carried out in Wells OXY-01, OXY 02, and POL-01 targeting the formation of interest, to find the key properties, namely, volume of shale, porosity, water saturation, permeability and hydrocarbon saturation. Two novel approaches have been used for the fracture characteristic analysis of the carbonate reservoir at Sakesar Limestone. The first practical approach analyzes the fracture network geometry at the crest of Balkassar anticline through an advanced Ant-tracking seismic attribute analysis. Subsequently, a proper Rock Physics Model (RPM) has been proposed on the basis of this initial estimate for a media with multiple fracture sets to study the spatial distribution of important fracture parameters (i.e., fracture density and their orientation) in the absence of sophisticated laboratory/wireline data. The second distinctive approach analyzes the conductive fracture by utilizing Formation Micro-Imager log (FMI) as a reference and utilizing machine learning AI algorithms. Initially, seismic data interpretation has been performed on available 3D seismic cube. The time events in terms of four horizons namely, Chorgali, Sakesar, Patala, and Lockhart have been marked on the seismic data based on synthetic seismogram and the three reverse faults, namely, F1, F2, and F3 have been marked based on the discontinuities across the traces and geological history of the Balkassar field. Sakesar limestone of Eocene age is zone of interest for the case study. To obtain an initial estimate regarding the presence of fracture corridors and their orientation in the formation of interest, the Ant-tracking discrete fracture network (DFN) attribute has been applied on the 3D post-stack seismic data. Based on this initial estimate, a proper RPM has been proposed which uses inverse Gassmann relations, T-matrix approximation, and Brown and Korringa relations. The output from the developed RPM has been displayed in the form of 13 effective independent elastic stiffness DRSML QAU vii constants (monoclinic symmetry – representing media comprising of multiple fracture sets) as a function of fracture densities and azimuthal fracture orientations. The second portion of this dissertation focuses on the identification of conductive fractures, as an essential tool for improving fluid flow extraction in case of a limited (conventional) dataset. To accomplish this task, an advanced machine learning Multi-feed Neural Network (MFNN) algorithms has been utilized to establish a relationship between the Continuous Wavelet Transformation (CWT) traces (Real, Imaginary and Magnitude), Coherence attribute and predicted FMI log at the well location. After successful training, this relationship has been simulated throughout the CWT volume for accurately mapping the probability distribution of conductive fractures and their effective porosities. Structural interpretation depicts the pop up anticlinal structure with severe thrust faults confirms compressional tectonic regime in Balkassar Oil field. The demarcated fault polygons extend in the NE SW direction. The contour models shows that the anticline has gentle limbs on the SE side and steep limbs on the NW side. The petrophysical results identifies that the Sakesar Formation has a producing potential in the study area of Balkassar oil field. OXY-01 shows an average volume of shale 6.9 %, an average porosity is 4% and an average water saturation is 60%. Well OXY-02 with 6% an average shale volume, 3.4 % an average porosity, and 41% an average water saturation. Well POL-01 with 4.1% an average shale volume, 5 % an average porosity, and 27.7% an average water saturation. The plots developed from the proposed rock physics model indicates a clear decreasing trend in effective elastic stiffness constants with increasing fracture densities. Similarly, a periodic trend of effective elastic stiffness constants with fracture orientations can be observed. These trends are more or less expected, but they would have been difficult to quantify without a proper rock physics model. The use of independent effective elastic constants for the generation of synthetic seismic amplitude versus angle and azimuth (AVAZ) data and its correlation with observed seismic AVAZ data in a geostatistical sense has been proposed. Coherency maps indicates subtle structural and stratigraphic features mainly faults, fractures and channels having low values highlighting fractures and other subtle patterns. The accuracy of the FMI log test was above 60 % making it reliable to be used for prediction at other available wells. Subsequently, the FMI logs for the Wells OXY-01 and OXY-02 have been predicted successfully with high accuracy validated by porosity logs and production of the well. The FMI log classified fractures into four classes, namely, conductive, discontinuous conductive, beddings, and clean carbonates (no fractures). CWT interpreted seismic signal in the form of real, imaginary and magnitude seismic traces. MFNN relation predicted the probability distribution of conductive fractures and their effective porosities. The red highlighted areas in the maps showing maximum probability of conductive fractures and maximum of 7% fracture porosity showing on porosity maps. The study provides an improved fracture mapping techniques that must have a significant impact on finding potential closure locations, fluid flow corridors, and areas with higher fracture permeability in tight carbonate reservoir