Monte Carlo-based approach for reservoir characterization

In this talk Dr. Shiv Shankar Ganguly presents his research titled: "A Bayesian multivariate model using Hamiltonian Monte Carlo inference to estimate total organic carbon content in shale "

This research paper focuses on developing a Hamiltonian Monte Carlo-based approach for reservoir characterization, specifically for estimating the total organic carbon (TOC) content in unconventional reservoirs. The study aims to address the challenges in characterizing these reservoirs due to the sparse TOC data available, which is crucial for capturing the hydrocarbon potential.  

Proposed Approach and Comparison with Conventional Methods The proposed approach combines geophysical well logs with the Hamiltonian
Monte Carlo method to estimate the posterior probability density of the model
parameters, including TOC content. The Bayesian framework allows for uncertainty
quantification in the model parameter estimation. The study compares the
performance of the approach with conventional empirical and physical law-based methods.

Performance Evaluation on Benchmark Data Using benchmark data from the Devonian Duvernay formation in Western Canada, the approach outperformed the traditional methods in terms of prediction accuracy. The mean absolute error, root mean square error, and correlation coefficient indicated that the Bayesian approach had better accuracy and reliability in estimating TOC content.

Validation with Real Field Data and Spatial Estimation The study also validated the approach using real field data from Silurian and other sources. The posterior probability distribution obtained from the simulations showed a close match with the reference TOC data, providing confidence in the model's accuracy. The study emphasized the importance of incorporating geophysical data, such as seismic data, to estimate TOC content in a 3D spatial manner. The researchers acknowledged the potential for further improvements in the approach, such as incorporating machine learning techniques. However, they highlighted the challenge of limited data availability for training such models in reservoir characterization. 
The paper concluded by recognizing the need to integrate advanced methods, such as the proposed Bayesian approach, in commercial software packages to enhance accessibility and applicability in the oil and gas industry.

Future Research Opportunities and Conclusion In terms of future research, the author mentioned a focus on Reservoir Characterization, both in conventional and unconventional resources, along with exploration of AI and machine learning approaches. The author also discussed the growing importance of clean energy and the role of geoscience in terms of mineral resource exploration, particularly for technologies like electric vehicles and solar PV plants. The paper encouraged researchers and young scientists to embrace these future challenges and opportunities in geoscience.

Paper reference: 
Ganguli, S.S., Kadri, M.M., Debnath, A., 2022. A Bayesian multivariate model using Hamiltonian Monte Carlo inference to estimate total organic carbon content in shale.​ Geophysics, 87(5), M163–M177.