85th EAGE Annual Conference & Exhibition

This paper investigates identification methods for weakly reflective reservoirs located beneath strong shielding horizons. The study focuses on two commonly used techniques: multi-wavelet decomposition and the reconstruction method based on eigenvalue decomposition in the frequency domain, known as matching pursuit. Through case studies using actual seismic data from the Dongying Depression, the effectiveness of these methods is demonstrated in enhancing the visibility of weakly reflective reservoirs and improving their correlation with other well-log curves. The study also explores the application of the generalized S-transform and waveform shaping techniques in the seismic data processing workflow. The findings contribute to advancing the understanding and utilization of these methods for improved reservoir characterization and exploration in areas with strong shielding horizons.

Real-time data analysis on drill cuttings images can greatly improve the efficiency of drilling operations. A low-power computation platform is preferred for this task because it is a challenge to provide reliable and stable power supplies directly in the field. Raspberry Pi is a popular edge-computing platform. Due to its low-power consumption and versatile computation capabilities, it has powered plenty of realistic real-time edge-computing applications. However, its low I/O throughput is a grand challenge for acquiring high-resolution drill cuttings images efficiently. One solution to effectively increase its I/O throughput is to only acquire a sparse portion of the original drill cuttings image and reconstruct it at a later stage when it is no longer time sensitive. We demonstrate the feasibility of using a Convergent POCS (CP) method running on a Raspberry Pi 4B device for reconstructing sparse images of drill cuttings. Our work paves the way to further exploit high-performance edge-computing solutions for automatically analyzing drill cuttings in real time.

The Sichuan Basin, located in the southeastern margin of the plateau, is controlled by the fracturing of the peripheral block and tectonic movement of the basement. During the Neoproterozoic period, the Sichuan Basin belonged to the Yangtze platform, which developed over a rifted continental margin that initiated along the southeastern side of the Yangtze block at ∼800 Ma ( Condon et al., 2005 ). The lithology of the Deng-1 Formation is mainly dolomite without fungus and algae, that of the Deng-2 Formation is algae-rich dolomite with snowflake-shaped structures and microorganisms with no snowflake-shaped structures, that of the Deng-3 Formation is mainly mudstone and classics, and that of the Deng-4 Formation is mainly algal dolomite and grey-black dolomitic mudstone.

For field development and drilling decisions, production assets and reservoir engineers require dynamic reservoir properties. Here we present a new method to estimate saturation (in fractions) and pressure (in bars) from time-lapse seismic data to provide to reservoir engineers. This new three-step method is demonstrated over the Edvard Grieg field in the North Sea. We can realize this method thanks accurate estimation of AVO parameters (P-impedance and S-impedance) from joint inversion of PP and PS ocean bottom data. The first step in our method is to estimate a stable set of axes identifying water saturation increase, gas saturation increase, pressure increase, and pressure decrease in the 4D AVO cross-plot domain. In the second step, using these axes, we convert every 4D P-impedance and S-impedance data point into 4D pseudo-saturation and pseudo-pressure data point using a projection method. In the third step, we use the rock physics relationships and reference values at key field location to map 4D changes in the field from impedances to actual saturation and pressure values. This allows us to obtain fieldwide dynamic values for water and gas saturation changes, and injection and production related pressure changes. These dynamic changes are interpreted over the field.

In this paper, we propose to use a bidirectional long short-term memory network with an attention mechanism (BiLSTM-Attention) to describe the spatial structure of 3D complex thin sand bodies. The 3D geological model contains a thin layer of river channels and alluvial fans with specific geological implications. We conduct high-resolution impedance inversion tests based on the sparse spike inversion, geostatistics inversion and BiLSTM-Attention network, respectively. On the evaluation of the inversion results, in addition to the presentation of the inversion profiles, we also perform a comparative analysis of the planar spreading of the sand body. The accuracy and precision of different methods for spatial spreading and distribution range of complex 3D thin sand bodies are compared comprehensively. The test results demonstrate that our proposed technical scheme achieves better inversion effects in terms of both profile and slice comparisons. The distinction of sand superposition relationships and the description of spatial spreading are more accurately.

The challenge of the nonlinear beamforming (NLBF) technology is to efficiently estimate local traveltime operators from input data, which has to solve millions or even billions of highly nonlinear optimization problems per data gather. The 2+2+1 method is a popular solver for estimating local traveltime operators, but its efficiency is not ideal on the CPU platform. We recently proposed a 2+2+1 GPU algorithm to accelerate its calculation performance by exploiting GPU calculation, and this 2+2+1 GPU implementation used the global memory of GPU as the primary work space because the data aperture engaged in NLBF may be too large so that only the global memory is spacious enough to hold the required work data. However, since it is also common for NLBF to use a small data aperture, it inspires us to propose an optimized GPU kernel for the 2+2+1 method with the shared memory of GPU as the primary work space. Furthermore, we also adopt Horner’s method and a GPU-friendly data reduction method to further improve the efficiency of the 2+2+1 GPU kernel. Our test on one field dataset shows an obvious efficiency improvement from this optimized 2+2+1 GPU kernel over the previous one.

Carbonate cave reservoirs, controlled by large strike-slip faults, hold significant potential for oil and gas exploration and development. Identifying these reservoirs is crucial but often challenging due to low SNR and the strong background reflection shielding in raw seismic data. To enhance seismic interpretation accuracy, we innovatively introduce the wavelet reconstruction and the weighted robust principal component analysis (WRPCA) techniques to eliminate strong background and noise interference. While the initial wavelet reconstruction can reduce part of the noise and background reflections, it falls short of fully separating reservoir reflections from interference signals. Consequently, the WRPCA technique is introduced post wavelet reconstruction to further achieve this goal. In comparison to traditional RPCA, WRPCA demonstrates higher accuracy and greater noise resistance, effectively isolating target reflections from background interference and noise. Subsequent attribute extraction and optimization are performed on the processed seismic data, with the instantaneous amplitude attribute chosen for cave reservoir identification. Applied to a basin in North China, these methods produce promising results, demonstrating their feasibility and validity in characterizing carbonate cave reservoirs.

The Least-squares migration can produce a high-quality migration images, but it is difficult to apply to the pre-processing stage due to the expensive computational cost. While the high-dimensional-space deblurring processing methods based on iconology are computationally efficient and can produce a high-resolution image, but have not been applied in the geophysics. How to combine the Least-squares migration theory with the deblurring technology of iconology is helpful to an efficient and high-resolution processing method. As we know, the migration image is the convolution of point spread function (PSF) and true reflection coefficient. The high-resolution processing method is actually an inverse problem of this forward problem. In this paper, a purely data-driven PSF extraction method is proposed based on the iconology. Then a high-dimensional space deconvolution algorithm is used to obtain high-resolution imaging results. The tests of data shows the proposed algorithm has the advantage of improving the resolution of geological bodies with different slope compared with the conventional high-resolution algorithms.

Field on production since 75 years, and sharing method to mitigate water challenges increase and risk of losing production

Azimuthal acoustic reflection imaging logging, which can measure the distance and azimuth of geological bodies beside the borehole within a range of tens of meters, is becoming an important technology for geophysical applications. Measurements for single-well acoustic reflection imaging in the model with a 1:1 reflection interface outside the borehole provide physical validation of this technique. In this paper, the analytical solution for the monopole reflection wavefield in the fluid-filled borehole is deduced by using the virtual source analogy. The simulation shows that the radiation wave from the nearby borehole is equivalent to the reflected wave from the reflector outside the borehole. On this basis, the experimental idea of using two parallel wells to achieve the verification experiment of single-well azimuthal reflected P-wave imaging is proposed. The experimental results show that the amplitude and arrival time of the azimuthal reflected P wave are sensitive to the azimuth of the reflector outside the borehole. This information can be used for precise positioning of geological anomalies outside the borehole. Under the guidance of theoretical simulations and experiments, the prototype of azimuthal reflected P-wave imaging logging tool was designed independently. Field tests were conducted, recording high signal-to-noise ratio azimuthal reflected P wave.