86th EAGE Annual Conference & Exhibition

CBM is an unconventional form of energy associated with coal as a by-product, which is developed during its formation and has composition similar to natural gas containing mostly methane. Due to lower carbon dioxide emission and high heat value on combustion, it is a superior fuel compared to coal and considered as a promising source of unconventional energy. India has the world’s fifth largest coal deposits, where its economy is also primarily dependent on coal for energy. The Indian coalfields consisted of Gondwana sediments, which are the potential storehouse of Coalbed Methane. Till date 16613km2 area has been identified for CBM exploration having prognosticated resource of about 62.4Tf3. Indian coal industries have highly committed towards CBM exploration, where around 9.9Tf3 CBM reserve has been explored as Gas in Place (GIP). The gas content of coal seams depends on several factors including its quality, occurrence depth, insitu conditions (temperature and pressure), maturity and permeability. The Indian coalfields mostly consists of bituminous to sub-bituminous coal, which are considered to potential storehouse of CBM. The present paper aims to study the dependence of gas content of major coal seams of Sohagpur Coalfield with respect to its quality, maturity and occurrence depth.

Coal seams are often associated with trapped methane and some of major Indian coalfields including Sohagpur Coalfield, Jharia Coalfield and Sonhat Coalfield are potential storehouses of CBM. On the basis of methane concentration and its emission, the DGMS has identified the Indian coal seams as Degree I, Degree II and Degree III category, where Degree III seams have highest methane content. In India, around 342 coal mines are currently operational in gassy coal seams, where 95 coal mines are working in the Degree II and Degree III coal seams. Methane is a highly inflammable gas and its concentration in underground mines must be less than 5% for avoiding gas explosions. Hence, degasification is very essential for reducing the risks of gas contamination and explosion for enhancing safety in coal mines. Hydrofracturing is a perforation method used for enhancing the CBM recovery from lower permeable coal seams. The effect of hydrofracturing on coal seams depends on its rank, cleat spacing, fracture density etc. which effects the penetration depth in a wide variation (2.5–60m). The present paper aims to understand the physical as well as mechanical changes due to hydrofracturing in the low permeable high rank bituminous coal seams of Sohagpur Coalfield.

The Lower Acacus reservoir is characterized by frequent alternation of sandstone, siltstone and shale. The vertical association of these rocks highly influences the reservoir quality and its lateral continuity. Therefore, the reservoir succession shows variation in the depositional pattern sequence from coarsening to fining upwards, as well as changes in the sedimentary structures and consequently disparity in the nature of paleo-flow. Integration of core image and FMI analysis assisted in the recognition of four main facies within the Lower Acacus reservoir; these are heterolithic bedding, cross-bedded sandstone, shaly-sand, and laminated shales. The association of these facies reflects a clear tidal effect on the deltaic deposition as supported by paleocurrent direction analysis. Moreover, the reservoir lithology is dominated by an argillaceous sandstone interbedded with shale and siltstone beds. The presence of clays and conductive minerals in the sand units, along with thinly laminated sand-shale units, creates difficulty in the hydrocarbon evaluation mainly because of the effect of those materials on resistivity measurements. This phenomenon is known as the Law Resistivity Pay (LRP), which is clearly exhibited by Lower Acacus shaly-sand reservoirs in the Ghadamis Basin.

Physical models are scaled representations of geological systems that enable greater understanding of structural deformation including salt tectonics. These models have been integrated into geophysical workflows for the creation of synthetic seismic forward modelling and imaging, by extraction of the physical model details and incorporation within an elastic earth model. This is achieved by utilizing digital photo imagery of model vertical cross-sections and using the pixel intensities and layer colours to provide detailed contrasts in seismic impedance. Seismic modelling and imaging can then be applied at field scales. This workflow is demonstrated on three case examples ranging from basin, prospect to reservoir scales. The resultant seismic data contains all the wealth of detail that is present in the original physical models and represents a realistic environment for testing of geophysical workflows including, modelling, imaging, processing, acquisition survey design and quantitative interpretation.

Seismic facies classification is generally performed in the domain of elastic attributes, such as P-impedance and P- and S- wave velocity ratio. The standard reservoir characterization workflows compute elastic properties from seismic data using seismic inversion methods based on Aki-Richards approximation, calculate the attributes of interest, and perform the facies classification in the selected domain. In this work, we propose a new AVO approximation formulated in the domain of the elastic properties used for the facies classification. We present a two-term AVO approximation in terms of P-impedance and P- and S- wave velocity ratio and a three-term AVO approximation in terms of P-impedance, P- and S- wave velocity ratio, and density. The formulations can be derived by combining the traditional formulations with rock physics relations, such as Gardner’s equation. We demonstrate the accuracy of the proposed equations by comparing them to the existing models for different reflectivity scenarios in two different geological settings.

Storing hydrogen in salt caverns at scale is one potential solution for both short and long-term storage of renewable energy sources, via transformation to and reuse of hydrogen as an energy carrier. To develop robust measurement, monitoring and verification planning for salt cavern storage, the detectability of salt inclusions and caverns must be assessed. However, many legacy seismic surveys have not been designed for detailed internal salt imaging and have source and receiver sampling that is too coarse. To assess the imageability with salt bodies a detailed numerical model has been constructed which has been constrained by both physical modelling and sonar profiles from actual gas storage caverns. Wave equation forward modelling and imaging has been applied to test the imaging limits and to provide guidance to monitoring schemes. In addition, the detection of engineered H2 migration out of the caverns is evaluated and quantified with 4D time-lapse seismic simulations. The results show that well sampled seismic surveys either from surface or downhole should be able to image these features at typical seismic frequencies in most cases, but that this will require higher trace densities that is available from legacy datasets.

To improve the accuracy of fracture prediction, we propose a fracture prediction method based on the fusion of pre-stack and post-stack seismic attributes. First, by analyzing the tectonic background, we establish fracture identification principles suitable for the geological context. Detailed structural interpretation is conducted based on post-stack seismic data volumes, leading to structural zoning and the analysis and systematic summary of seismic response characteristics. This forms a fundamental understanding of the spatial development patterns of fractures. On this basis, various seismic attribute volumes are extracted, including enhanced coherence volumes, maximum likelihood volumes, and curvature attribute volumes, with different sensitivity parameters set for different subareas. Pre-stack anisotropic inversion is performed using multi-azimuth AVO methods to extract pre-stack azimuthal anisotropy intensity and orientation data volumes. Through multi-attribute selection and fusion, fracture attribute identification is optimized, and weight parameters are calibrated based on actual fracture characteristics observed in wells, resulting in an accurate spatial distribution volume of fractures. This integrated method not only enhances the resolution of fracture prediction but also allows the identification of fractures of different scales and characteristics, providing a foundation for further reservoir studies and development.

Deep learning (DL) methods have shown significant potential in post-stack seismic impedance inversion, but their effectiveness is often limited by the availability of training data. Using forward modeling to enrich labeled datasets or embedding a forward modeling module within the network can help mitigate this issue. However, these approaches often rely on wavelets or increase training complexity. This paper proposes a dual-branch network architecture for seismic impedance inversion. The model combines multi-sized convolutional kernels for feature extraction and employs a self-attention mechanism to capture relationships between different sampling points. The results of local and global feature extraction are fused in real time, ensuring that detailed information is preserved while global patterns are learned. Additionally, the initial impedance model is used as an input to constrain the training process. This dual-branch structure enables the model to fully learn the nonlinear relationship between seismic data and impedance without relying on seismic wavelets or a forward modeling module. Experiments on synthetic and field data demonstrate that the proposed method accurately predicts impedance with limited samples, effectively delineating complex geological structures.

Full Wave Inversion (FWI) is an effective tool for estimating subsurface velocity models; however, its high nonlinearity presents certain limitations, including significant computational costs in terms of time and hardware resources. Accelerating the FWI inversion process without compromising performance remains a challenging task. To address this issue, we propose the FreqSiameseFWI framework, a deep learning-based misfit function designed to expedite the FWI inversion process and support Multi-source FWI (MSFWI) by mitigating the impact of cross-talk noise. FreqSiameseFWI employs a self-supervised learning approach integrated within the FWI framework, enabling iterative updates of its parameters without introducing significant additional overhead costs. The seismic data are converted to the frequency domain through the Fast Fourier Transform (FFT), allowing the Siamese network to extract spectral features from the seismic data and achieve robust inversion performance. The proposed FreqSiameseFWI effectively mitigates the influence of cross-talk noise. The performance of FreqSiameseFWI, evaluated using the Overthrust model, has yielded promising results that outperform conventional misfit functions.

Acoustic remote detection technology plays a pivotal role in the efficient exploration and development of fractured reservoirs. This technology significantly enhances the accuracy of identifying complex underground geological structures, bringing remarkable benefits to increasing oil and gas production. Nonetheless, the considerable computational demands and the inefficiency in the 3D elastic wave time-domain FD method have impeded extensive adoption within the realm of acoustic remote detection imaging. This study introduces a high-performance technique for three-dimensional elastic wave simulation, which employs a synchronous CPU+GPU heterogeneous parallelism architecture. An optimized dimensionality reduction strategy and 3D-Hbrid-PML absorption boundary have been implemented, significantly elevating the computational efficiency for 3D simulations, thereby enabling the simulation of large-scale models with fracture. The method’s stability, efficiency and precision were verified by conducting numerical simulations on the fracture model. The results indicate that the morphology and orientation of the fracture can be accurately identified by the simulation data.