Uso de redes adversárias geradoras condicionais para construção de modelos de velocidades sísmicas

Abstract

In the oil and gas industry, estimating an accurate seismic velocity model is an essential step in seismic processing, reservoir characterization, and hydrocarbon volume calculation. Full Waveform Inversion (FWI) velocity modeling is an iterative advanced technique that provides a high-resolution seismic interval velocity model, although at a very high computational cost due to the wave-equation-based numerical simulations required at each FWI iteration. This study proposes a method of generating seismic velocity models, as detailed as those obtained through FWI, however using a fraction of the time and computational resources spent in traditional FWI velocity modeling. To do so, a 3D conditional Generative Adversarial Network (cGAN), known as pix2pix was used. The cGAN training was performed using three conditional inputs, Seismic Amplitude, Tomografic Average Velocity, and Two Way Time Grid, the desired output was the high-resolution seismic interval velocity model from the FWI modeling. Real-world data were used to train and validate the proposed network architecture, and three evaluation metrics (percent error, structural similarity index measure, and visual analysis) were adopted as quality criteria. Based on these metrics, the results evaluated upon the validation set suggest that the cGAN was able to accurately match real FWI generated outputs, enabling it to extract from input data the main geological structures and lateral velocity variations. Experimental results indicate that the proposed method, when deployed, has the potential to increase the speed of geophysical reservoir characterization processes, saving time and computational resources.