Nonstationary seismic reflectivity inversion based on prior-engaged semisupervised deep learning method

Abstract

Reflectivity inversion methods based on a stationary convolution model are essential for seismic data processing. They compress the seismic wavelet, and by broadening the bandwidth of seismic data, they assist the interpretation of seismic sections. Unfortunately, they do not apply to realistic nonstationary deconvolution cases where the seismic wavelet varies as it propagates in the subsurface. Deep learning techniques have been proposed to solve inverse problems where networks can behave as the regularizer of the inverse problem. Our goal is to adopt a semi-supervised deep learning approach to invert reflectivity when the propagating wavelet is considered unknown and time-variant. To this end, we design a prior-engaged neural network by unrolling an alternating iterative optimization algorithm, where convolutional neural networks are used to solve two sub-problems. One is to invert the reflectivity, and the other is to estimate the time-varying wavelets. Generally, it is well-known that when working with geophysical inverse problems such as the one at hand, one has limited access to labeled data for training the network. We circumvent the problem by training the network via a data-consistency cost function where seismic traces are honored. Reflectivity estimates are also honored at spatial coordinates where true reflectivity series derived from borehole data are available. The cost function also penalizes time-varying wavelets from varying abruptly along the spatial direction. Experiments are conducted to show the effectiveness of the proposed method. We also compared the proposed approach to a nonstationary blind deconvolution algorithm based on regularized inversion. Our findings show that the proposed method improves the vertical resolution of seismic sections with noticeable correlation coefficient improvements over the nonstationary blind deconvolution. In addition, the proposed method is less sensitive to initial estimates of nonstationary wavelets. Moreover, it needs less human intervention when setting parameters than regularized inversion.

High‐resolution reservoir prediction method based on data‐driven and model‐based approaches

“…Chai et al (2021) used convolutional neural networks (CNNs) to achieve sparse-spike deconvolution, reducing the impact of previous human-computer interactions. Chen et al (2023) implemented deconvolution using a semi-supervised DL method, and the correlation coefficient is higher than nonstationary blind deconvolution. Dong et al (2014) implemented a processing from low-resolution images to high-resolution images using CNNs and pointed out that the method has strong generalization and stability.…”

“…Chen et al. (2023) implemented deconvolution using a semi‐supervised DL method, and the correlation coefficient is higher than nonstationary blind deconvolution. Dong et al.…”

Amplitude Variation with Offset (AVO) Inversion for Reservoir Visualization: A Case Study of Taje Field, Niger Delta, Nigeria

“…Seismic reflectivity inversion and interpretation have heavily relied on Acoustic Impedance (AI) since its introduction in the 1970s. AI is valued for its strong agreement with rock attributes measured in experiments and field data [6] . Unlike seismic reflectivity, which occurs at the interfaces of different strata, AI values remain constant within rock layers, simplifying the link with geology and stratigraphy.…”

“…AI transformations are crucial for seismic interpretation and reservoir characterization, particularly in identifying fluid-filled and porous zones [9] . Typically, stacked seismic data are used to estimate normalincidence reflectivity, from which AI is inverted [6] .…”

Deep decomposition learning for reflectivity inversion

“…While a more realistic model entails an unknown and non‐stationary propagating wavelet, this simplification serves as a good approximation. We point out, however, that multiple efforts have also considered time‐variant and blind deconvolution frameworks that simultaneously estimate the wavelet and the reflectivity in a non‐linear fashion (e.g., Chen et al., 2023; Gholami & Sacchi, 2013; Kaaresen & Taxt, 1998; Kazemi & Sacchi, 2013).…”