EAGE Workshop on Advanced Seismic Solutions for Complex Reservoir Challenges

Spectral decomposition, or time-frequency analysis, is an advanced seismic attribute analysis technique that decomposes seismic data into its frequency components, providing different insights into subsurface geology. This study examines the principles, methodologies, and applications of spectral decomposition, including methods such as Short-Time Fourier Transform (STFT), Continuous Wavelet Transform (CWT), and high-resolution techniques like Basis Pursuit (BP) and Empirical Mode Decomposition (EMD), to address varying resolution requirements in seismic data analysis. Emphasis is placed on detecting hydrocarbon reservoirs using amplitude variation with frequency (AVF) analysis and seismic attenuation phenomena, delineating thin beds, and enhancing seismic visualization. The AVF analysis emerges as a simple and handy tool for identifying hydrocarbon presence through frequency-dependent attenuation, even in the absence of pre-stack seismic data, with attenuation phenomena linked to lithology and fluid interactions offering valuable insights for reducing exploration risks. Through case studies, the spectral decomposition with AVF analysis is demonstrated, showcasing its effectiveness in delineating subsurface features, mitigating interpretation pitfalls, and accurately detecting hydrocarbons. This work highlights the transformative role of spectral decomposition in seismic interpretation, blending diverse perspectives, analytical precision, and innovative visualization methods to advance geological exploration.

This paper investigates the impact of near-surface geological complexities on seismic imaging challenges encountered at deeper reservoir levels. It highlights how heterogeneity within near¬surface layers significantly influences the accuracy and effectiveness of deeper seismic interpretations. Through a comprehensive analysis of various onshore and offshore 3D seismic surveys, the paper discusses strategies employed to address these near-surface imaging obstacles. The findings underscore the importance of understanding near-surface dynamics in enhancing seismic imaging techniques, obtaining more accurate subsurface models and demonstrate improvements in the clarity and reliability of seismic data at deeper reservoir levels. We develop a refined velocity model that enhances the understanding of the geometrical features. Through advanced modeling techniques, we analyse how these geological features influence seismic wave propagation, thereby improving imaging accuracy. The findings provide critical insights into the subsurface characteristics associated with the complex geology, facilitating better exploration and characterization of underlying geological formations. We elaborate on fundamental full waveform inversion quality control steps and methodologies. The examples are showing improvement in structural definition, signal to noise ratio and reflection continuity with preserved relative amplitudes.

The travel time of shear waves (DTS) in caprock and carbonate formations is critical for subsurface studies, including evaluating caprock sealing integrity, storage capacity, lithology, and geomechanical modeling. However, DTS logs are often unavailable in older wellbores due to data loss or lack of recording. This study develops a mathematical model to estimate DTS from compressional sonic logs (DTC) using a simple supervised machine learning algorithm, Random Forest, for calibration and verification.

The study utilizes geophysical logs such as gamma ray (GR), neutron porosity (NPhi), and density (RhoB) for predictions. Case studies are conducted on two carbonate fields in the Luconia Basin, offshore Sarawak, using offset well data from Field-A. Separate models correlating DTC and DTS are generated for caprock and carbonate reservoirs and applied to other wells in Field-A and Field-B to create synthetic DTS logs.

Machine learning independently calibrates the synthetic DTS using well log datasets, with consistent results observed between the mathematical model and machine learning outputs. Comparing linear regression with machine learning reveals strong agreement, with the model showing the highest R² value considered most reliable for detailed subsurface studies. This approach demonstrates the feasibility of DTS estimation using readily available log data.

Capillary pressure (Pc) data is utilized to calculate saturations through saturation height (SHF) modelling in a carbonate field, calibrated to an empirical model. Hydrocarbon saturation is the fraction of pore volume filled with gas, and this is crucial for reserve estimations. Water saturation (Sw) can be estimated via core analysis of Dean-Stark extractions, empirical models using logs data and SHF modelling. Archie’s equation calculates Sw from well logs, while Leverett’s equation calculates Sw-SHF. The latter is a reliable predictor of Sw, independent of the electrical properties in heterogenous carbonate environment. This method reduces the uncertainties linked to formation salinity and proved to be highly accurate.

Fluid contacts are defined by integrating data from repeat formation tester (RFT), drill stem test (DST), production logging tool (PLT), pressure volume temperature (PVT) sample analysis, and petrophysical well logs. Alternatively, the SHF modelling also verified the contacts. This SHF approach achieves excellent saturation matches at the wells level and used for field-wide implementations.

With an axisymmetric connected fracture model, this paper discovers that the resonance of acoustic waves within the connected fractures yields reflected Stoneley waves of many oscillation cycles. The inherent frequency of the resonant waves changes with the extension length of the fracture and with the depth of mud invasion for gas layer. Stemming from these discoveries, this paper proposes a method for identifying connected fractures and gas layers using the attenuation coefficient of Stoneley waves

The low-frequency attenuation coefficient anomalies of the pulsed Stoneley wave in the borehole can be used to characterize the properties of the formations, guide lateral drilling, and provide information to large scale hydraulic fracturing. This is one of the techniques to increase deep well exploration efficiency. The amplitude of Stoneley wave reaches to the maximum at the borehole wall. It then attenuates gradually from there externally. The lower the frequency, the slower the attenuation. The wavefront exhibits olive-shaped when it penetrates deep into the formation. Attenuation coefficient anomalies show up when the wavefront encounters interfaces or fractures at certain low frequencies. Those low frequency anomalies can be used to characterize the interfaces and fractures at different radial depths. The attenuation coefficient spectrum described in this paper is different from the Stoneley wave attenuation coefficient calculated directly from amplitude. The attenuation coefficient spectrum carries the information about the radial profile of the formation.

STC method is a traditional method used to obtain P-wave and S-wave slowness of the formation from array waveforms. This method is valid only when the slowness dispersion is low. This paper develop a method to obtain slowness of P-wave and S-wave using slowness dispersion curve, which is little affected by the influence of slowness dispersion.

Within the range where frequency is higher than the natural frequency of compressional wave of formation, slowness of slowness dispersion curve is closest to the slowness of compressional wave. Within the range where cutoff frequency is included, the slowness of slowness dispersion curve is closest to the slowness of shear wave. Projecting the slowness dispersion curve onto the slowness axis, we obtain a curve varying with slowness, On this curve, each slowness correspond to a value which peak at the slowness of compressional wave and slowness of shear wave.

This study explores the application of Artificial Intelligence and Machine Learning (AIML) in interpreting overburden faults within Carbon Capture and Storage (CCS) studies. By employing a progressive adaptive model training approach, the research aims to achieve efficient and precise fault prediction. The results demonstrate that the AIML system significantly enhances fault location precision, identifies previously undetected faults, and provides valuable insights into potential hydrocarbon migration pathways. The study underscores the critical importance of detailed overburden structural and fault seal analysis for ensuring safe CO2 containment. Additionally, it contributes novel information on risk zones for CO2 storage in depleted oil and gas reservoirs, highlighting the effectiveness of AIML technology in improving fault location accuracy.

Seismic interpretation is vital for subsurface resource exploration, including CO2 storage projects. Traditional workflows rely on bulk-shifted horizons, often misaligning with seismic reflectors, leading to suboptimal reservoir characterization. The surgical mapping workflow enhances accuracy by systematically reviewing data, conditioning it, and precisely mapping minor reservoirs before extracting seismic attributes. Tested on Field A offshore Peninsular Malaysia, it revealed significant discrepancies between bulk-shifted and surgically mapped horizons, especially near faults. RMS amplitude analysis showed clearer geological features, by correlating with well logs and depositional environments. This method provides improved precision in CO2 storage evaluation, injector well placement, plume migration modelling, and MMV planning. It ensures better-informed decision-making, optimized well placement, and enhanced project performance. Additionally, it can aid in assessing top seal formations for CO2 containment security. The surgical mapping workflow is a valuable tool for accurate subsurface characterization, contributing to safer and more effective CCS project execution.