Oil & Gas

This paper presents an AI-driven workflow designed for the automatic interpretation of large seismic volumes, applied to a 15,000 km² offshore area in Abu Dhabi. The workflow begins with fault detection for low-magnitude, strike-slip faults, achieved through fine-tuning deep neural networks (DNNs). This fault detection process is enhanced by structure-preserving denoising to maintain critical features. Next, automated horizon interpretation and flattening reveal a variety of geological features that are otherwise difficult to identify. Finally, structure-enhancing denoising improves geological feature detection, using channel detection as an example. This workflow is set to be expanded to cover the entire offshore Abu Dhabi region, offering significant potential for large-scale geological analysis.

In this article, we explore how advanced imaging techniques are enhancing the characterization of hydrocarbon reservoirs in the Barents Sea. Bright amplitudes on seismic data beneath the Base Quaternary often signal potential hydrocarbon accumulations, but accurately interpreting these signals requires in-depth analysis of amplitude versus reflection angle (AVA) information. Traditional seismic imaging methods, such as Q-Kirchhoff pre-stack depth migration (QPSDM), may be limited by shallow gas pockets, surface-related multiples, and other distortions. However, full-waveform inversion (FWI) is changing the game by providing clearer, more accurate reflectivity images without the need for extensive pre-processing or migration. This study demonstrates how extracting elastic data from acoustic FWI images can yield AVA products comparable to those from conventional methods, but with the added advantage of cleaner and more accurate results, especially in areas with complex near-surface geology.