It is of great significance to detect faults correctly in continental sandstone reservoirs in the east of China to understand the distribution of remaining structural reservoirs for more efficient development operation. However, the majority of the faults is characterized by small displacements and unclear components, which makes it hard to recognize them in seismic data via traditional methods. We consider fault detection as an end-to-end binary image-segmentation problem of labeling a 3D seismic image with ones as faults and zeros elsewhere. Thus, we developed a fully convolutional network (FCN) based method to fault segmentation and used the synthetic seismic data to generate an accurate and sufficient training data set. The architecture of FCN is a modified version of the VGGNet (A convolutional neural network was named by Visual Geometry Group). Transforming fully connected layers into convolution layers enables a classification net to create a heatmap. Adding the deconvolution layers produces an efficient network for end-to-end dense learning. Herein, we took advantage of the fact that a fault binary image is highly biased with mostly zeros but only very limited ones on the faults. A balanced crossentropy loss function was defined to adjust the imbalance for optimizing the parameters of our FCN model. Ultimately, the FCN model was applied on real field data to propose that our FCN model can outperform conventional methods in fault predictions from seismic images in a more accurate and efficient manner.

Anelasticity of the earth subsurface medium, which is quantified by the quality factor $Q$ , causes the dissipation of seismic energy. Strong attenuation effect resulting from geology such as gas clouds (gas-filled sandstone) is a challenging problem for high-resolution imaging. To compensate the attenuation effect, first we need to accurately estimate the attenuation parameter. However, it is difficult to directly derive a heterogeneous attenuation Q model. This research letter proposes a method to derive a Q model corresponding to strong attenuative media from marine reflection seismic data using convolutional neural network (CNN), a popular deep learning framework. We treat Q anomaly detection problem as a semantic segmentation task and train a network to perform a pixel-by-pixel prediction to invert a pixel group that belongs to the strong level of attenuation probability. The proposed method uses a volume of marine 3-D reflection seismic data for network training and validation, which needs only a small part of real data as the training set due to the feature of U-Net. In the final stage, to evaluate the attenuation model, we validate the predicted heterogeneous Q model using deabsorption prestack depth migration (Q-PSDM), a high-resolution imaging result in depth domain with appropriate compensation is obtained.

A high-resolution seismic image is the key factor for helping geophysicists and geologists to recognize the geological structures below the subsurface. More and more complex geology has challenged traditional techniques and resulted in a need for more powerful denoising methodologies. The deep learning technique has shown its effectiveness in many different types of tasks. In this work, we used a conditional generative adversarial network (CGAN), which is a special type of deep neural network, to conduct the seismic image denoising process. We considered the denoising task as an image-to-image translation problem, which transfers a raw seismic image with multiple types of noise into a reflectivity-like image without noise. We used several seismic models with complex geology to train the CGAN. In this experiment, the CGAN’s performance was promising. The trained CGAN could maintain the structure of the image undistorted while suppressing multiple types of noise.

Active fault detection has an important significance for seismic disaster prevention and mitigation in urban areas. The high-density station arrays have the potential to provide a microtremor survey solution for shallow seismic investigations. However, the resolution limitation of the nodal seismometer and small-scale lateral velocity being inhomogeneous hinder their application in near-surface active fault exploration. Distributed acoustic sensing (DAS) has been developed rapidly in the past few years; it takes an optical fiber as the sensing medium and signal transmission medium, which can continuously detect vibration over long distances with high spatial resolution and low cost. This paper tried to address the issue of near-surface active fault exploration by using DAS. We selected a normal fault in the southern Datong basin, a graben basin in the Shanxi rift system in north China, to carry out the research. Microtremor surveys across the possible range of the active fault were conducted using DAS and nodal seismometers, so as to obtain a shallow shear wave velocity model. Meanwhile, we applied a Brillouin optical time domain reflectometer (BOTDR) and distributed temperature sensing (DTS) to monitor the real-time fluctuation of ground temperature and strain. Our results show that the resolution of the deep structures of the fault via the microtremor survey based on DAS is lower than that via the seismic reflection; whereas, their fault location is consistent, and the near-surface structure of the fault can be traced in the DAS results. In addition, both the BOTDR and DTS results indicate an apparent consistent change in ground temperature and strain across the fault determined by the DAS result, and the combination of surface monitoring and underground exploration will help to accurately avoid active faults and seismic potential assessment in urban areas.

The Jidong Oilfield No. 2 Underground Gas Storage (UGS), located in an active fault zone in Northern China, is a key facility for ensuring natural gas supply and peak regulation in the Beijing–Tianjin–Hebei region. To evaluate the effectiveness of a combined surface and shallow borehole monitoring system under deep reservoir conditions, a 90-day microseismic monitoring trial was conducted over a full injection cycle using 16 surface stations and 1 shallow borehole station. A total of 35 low-magnitude microseismic events were identified and located using beamforming techniques. Results show that event frequency correlates positively with wellhead pressure variations instead of the injection volume, suggesting that stress perturbations predominantly control microseismic triggering. Events were mainly concentrated near the bottom of injection wells, with an average location error of approximately 87.5 m and generally shallow focal depths, revealing limitations in vertical resolution. To enhance long-term monitoring performance, this study recommends deploying geophones closer to the reservoir, constructing a 3D velocity model, applying AI-based phase picking, expanding array coverage, and developing a microseismic-injection coupling early warning system. These findings provide technical guidance for the design and deployment of long-term monitoring systems for deep reservoir conversions into UGS facilities.