Exxonmobil Upstream Research Company

David H Johnston

Brian P West

John E Eastwood

Michael D Gross

Dwight C Dawson

Partha Sarathi Routh

Ganglin Chen

Musa Maharramov

Provided is a method for determining convergence in full wavefield inversion (FWI) of 4D seismic (time-lapse seismic: 3D seismic surveys acquired at different times with the first survey termed as the baseline and subsequent surveys termed as monitors). FWI applied to field seismic data includes iteratively solving for subsurface property models and model difference between monitor and baseline. Iteration occurs until the model difference is sufficiently converged. Rather than determining convergence by examining an entire subsurface region of the models and/or the model difference, subparts of the subsurface region models and/or the model difference are examined in order to determine convergence. For example, different regions behave differently, include the target reservoir region (where hydrocarbon is present) and the background region that is outside the target reservoir region. Thus, transforming the subregions of the models and/or the model difference and analyzing the transformations may indicate convergence of the overall model difference.

Method for determining convergence in full wavefield inversion of 4D seismic data

Near-offset and far-offset seismic data volumes are time-aligned by first selecting a plurality of time shifts. The near-offset and far-offset seismic data volumes are cross-correlated at the plurality of time shifts. An initial time-shift volume and a maximum correlation volume are created from the maximal cross-correlations at the plurality of time shifts. Areas of high time shift from the initial time-shift volume and areas of low cross-correlation from the maximum correlation volume are determined. The determined areas of high time shift and low cross-correlation are filtered from the initial time-shift volume, generating a filtered time-shift volume. The filtered time-shift volume is applied to the far-offset seismic volume to generate a time-aligned far-offset volume.

Method for time-aligning multiple offset seismic data volumes

AVO anomalies are classified in near-offset and far-offset seismic data volumes, by first calculating a plurality of initial AVO seismic attributes representative of the offset seismic data volumes. A probabilistic neural network is constructed from the calculated initial AVO seismic attributes. AVO anomaly classifications are calculated in a portion of the offset seismic data volumes. The preceding steps are repeated until the calculated AVO anomaly classifications in the portion of the offset seismic data volumes are satisfactory. AVO anomaly classifications are calculated throughout the offset seismic data volumes using the constructed probabilistic neural network.

Method for classifying AVO data using an interpreter-trained neural network

Growing community of inventors

Houston, TX, United States of America

David H Johnston