Full text loading...
-
Investigating the AVO Signature of a North Sea Oil-Field Using Pseudo-Wells and Unsupervised Deep Learning
- Publisher: European Association of Geoscientists & Engineers
- Source: Conference Proceedings, 81st EAGE Conference and Exhibition 2019, Jun 2019, Volume 2019, p.1 - 5
Abstract
In this study, a workflow for seismic lithology and fluid prediction is presented using rock physics based pseudowells and deep learning algorithms. The approach is referred to as ASAP (“Anomaly Screening Applying Pseudowells”) and provides a rapid method for quantitative seismic interpretation and prospect mapping in areas with limited well control. The method extends on conventional AVO classification, where only intercept and gradient information is used, by taking into account the full waveform in the AVO data. Convolutional neural network models are utilized to train and map out latent parameters from the seismic data. The calibration with a discovery well and the application of 240 pseudo-wells allow for an intuitive understanding of the latent space and rapid classification of hydrocarbon filled reservoirs away from well control. Pseudo-wells can be easily updated on-the-fly if alternative rock physics models or additional geological scenarios were to be tested.