Chasing Middle Cretaceous Stratigraphic Traps in Offshore Abu Dhabi by Incorporating 3D Seismic Data into Machine Learning Methodology

Abstract

The aim of this study is to explore the Middle Cretaceous hydrocarbon potential particularly associated with prograding units. A machine learning based methodology is applied to evaluate the exploration potential of the Cretaceous Mishrif clinoforms in offshore Abu Dhabi, where a recent high-quality 3D seismic data have been acquired. The exploration objectives are to perform structural interpretation of the play; to infer reservoir quality, such as porosity; and to characterize the internal geometry of the reservoir and its seismic lithofacies. Machine learning is a branch of artificial intelligence (AI) that plays a crucial role for evaluating data and building knowledge. In the context of subsurface characterization, the integration of AI into hydrocarbon exploration brings a substantial value for appraising different plays in a faster, robust and reliable manner. This is especially relevant in early exploration stages, when the availability of well data is limited but huge amount of high quality seismic data exist. Advantages of applying machine learning algorithms are its capability (1) to deal with several types of information and large 3D seismic data; (2) to learn by themselves from these data; and (3) to assist in interwell pre-drill estimates and associated uncertainties in a semi-automated context. Seismic inversion is a process that integrates both 3D seismic and well data to derive reservoir properties. The seismic derived elastic models can be incorporated in reservoir modelling stage as secondary data, indicating the variability of the properties where there is no well information. Moreover, assessing the uncertainty related to reservoir quality of the Mishrif play will allow more effective exploration decisions, such as well placement and volumetric estimations. Clinoforms depositional surfaces are identified in the seismic inverted elastic models, displaying lateral variation in geometry and reservoir properties. Given the high correlation achieved between acoustic impedance and log-porosity, it is infer with relative high confidence that better porosity development is expected to be found at the topmost part of the clinoforms. The quality of the reservoir is anticipated to degrade towards the edge of the clinoforms. Hydrocarbon charge model from the nearby Shilaif source rock kitchens will be incorporated to reduce the risk of future exploration activities in the area.