The goal of image recognition applied to seismic interpretation for hydrocarbon exploration is to identify the location of geological features that could be related to oil accumulation. The evaluation of structures on the subsurface is an important aspect since a large number of hydrocarbon reservoirs are contained in some kind of structural trap. Many types of structures are created by earth’s stresses and are called structural traps, among these are the anticlines. The anticline patterns range from simple to exceedingly complex, and its evaluation requires the development of new types of computational approaches to extract the useful and valuable underlying information for interpretation. To addresses these approaches, in this paper, we present a case study where a deep learning algorithm: Convolutional Neural Networks (CNN), is invoked to train a predictive model over 2D synthetic seismic images correspond to anticline structures containing gas and water, respectively. We show that the CNN algorithm has the ability to correctly infer and classify anticlines structures not used during the training process of the algorithm. We conclude that CNN is a promising mechanism to automatically detect and identify a high percentage of anticlines structures on seismic data.
Presentation Date: Wednesday, September 18, 2019
Session Start Time: 1:50 PM
Presentation Start Time: 3:05 PM
Presentation Type: Oral