Chapter 9 - Regularized elastic full-waveform inversion using deep learning

Abstract

Elastic full-waveform inversion, which aims to match the waveforms of prestack seismic data, potentially provides more accurate high-resolution reservoir characterization from seismic data. However, full-waveform inversion can easily fail to characterize deep-buried reservoirs due to illumination limitations. We present a deep learning-aided elastic full-waveform inversion strategy using observed seismic data and available well logs in the target area. Seismic facies interpreted from well logs are linked to the inverted P- and S-wave velocities using trained neural networks, corresponding to the subsurface facies distribution. The desired reservoir-related parameters such as velocities and anisotropy parameters are evaluated using a weighted summation given by the neural network classification distribution of facies. Finally, we update these estimated parameters by matching the resulting simulated wave fields to the observed seismic data, which corresponds to another round of elastic full-waveform inversion aided by the prior knowledge gained from ML predictions. A modified Marmousi synthetic example is used to prove the concept of the proposed inversion method. A North Sea field data example, the volve oil field data set, shows that the use of facies as prior helps resolve the deep-buried reservoir target better than the use of only seismic data.