Gan, Thomas (Shell Trinidad & Tobago Ltd) | Kumar, Ashok (Shell Trinidad & Tobago Ltd) | Ehiwario, Michael (Shell Exploration & Production Company) | Zhang, Barry (Quantico Energy Solutions) | Sembroski, Charles (Quantico Energy Solutions) | de Jesus, Orlando (Quantico Energy Solutions) | Hoffmann, Olivier (Quantico Energy Solutions) | Metwally, Yasser (Quantico Energy Solutions)
Borehole-log data acquisition accounts for a significant proportion of exploration, appraisal and field development costs. As part of Shell technical competitive scoping, there is an ambition to increase formation evaluation value of information by leveraging drilling and mudlogging data, which traditionally often used in petrophysical or reservoir modelling workflow.
Often data acquisition and formation evaluation for the shallow hole sections (or overburden) are incomplete. Logging-while-drilling (LWD) and/or wireline log data coverage is restricted to mostly GR, RES and mud log information and the quality of the logs varied depending on the vendor companies or year of the acquisition. In addition, reservoir characterization logs typically covered only the final few thousand feet of the wellbore thus preventing a full quantitative petrophysical, geomechanical, geological correlation and geophysical modelling, which caused limited understanding of overburden sections in the drilled locations and geohazards risls assessment.
Use of neural networks (NN) to predict logs is a well-known in Petrophysic discipline and has often used technology since more than last 10 years. However, the NN model seldon utilized the drilling and mudlogging data (due to lack of calibration and inconsistency) and up until now the industry usually used to predict a synthetic log or fill gaps in a log. With the collaboration between Shell and Quantico, the project team develops a plug-in based on a novel artificial intelligence (AI) logs workflow using neural-network to generate synthetic/AI logs from offset wells logs data, drilling and mudlogging data. The AI logs workflow is trialled in Shell Trinidad & Tobago and Gulf of Mexicooffshore fields.
The results of this study indicate the neural network model provides data comparable to that from conventional logging tools over the study area. When comparing the resulting synthetic logs with measured logs, the range of variance is within the expected variance of repeat runs of a conventional logging tool. Cross plots of synthetic versus measured logs indicate a high density of points centralized about the one-to-one line, indicating a robust model with no systematic biases. The QLog approach provides several potential benefits. These include a common framework for producing DTC, DTS, NEU and RHOB logs in one pass from a standard set of drilling, LWD and survey parameters. Since this framework ties together drilling, formation evaluation and geophysical data, the artificial intelligence enhances and possibly enables other petrophysical/QI/rock property analysis that including seismic inversion, high resolution logs, log QC/editing, real-time LWD, drilling optimization and others.
Water production in organic-rich mudrock formations is one of the most critical challenges associated with unconventional oil and gas development. However, the factors affecting water production is still not well understood. Flow and distribution of fluids in porous media are significantly influenced by dynamic petrophysical properties, which are affected by wettability. We have recently documented a significant impact of geochemistry on wettability of organic-rich mudrocks, which can potentially affect multi-phase fluid flow and water production in these reservoirs. Therefore, understanding the impacts of geochemistry, thermal maturity, and wettability on water production is crucial for production optimization in organic-rich mudrocks. The objective of this paper is to investigate the impact of geochemistry, thermal maturity, and wettability on water and hydrocarbon production in organic-rich mudrocks.
We first performed an experimental procedure to determine the effect of thermal maturity of mudrock on brine permeability of crushed samples. We crushed mudrock samples from an organic-rich mudrock formation, and synthetically matured these samples by heat-treating them. We then performed brine core-flood on the crushed samples at different thermal maturity levels filled in a specially designed core chamber. The experimental data indicated that increase in thermal maturity of the samples from Hydrogen Index (HI) of 328 to 54 mg hydrocarbon/g organic carbon (mg-HC/g-OC) increases the brine permeability from 125 to 550 mD. We then investigated the impacts of thermal maturity on water production in two wells drilled in an organic-rich mudrock formation. We first performed conventional well-log interpretation to obtain volumetric concentrations of minerals and petrophysical properties. We used the well-log-based estimates of petrophysical, geochemical, and compositional properties as inputs to well-log-based rock classification using unsupervised neural-network method. Analysis of production data in the same rock type with similar petrophysical and compositional data, revealed that rock types with higher thermal maturity have higher relative water production by up to 56% as compared to rock types with low thermal maturity. On the other hand, hydrocarbon production was higher by 83% in rock types with low thermal maturity as compared to low thermal maturity rock types. This could be explained by the increase of brine permeability with increase in thermal maturity due to differences in wettability of the two rock types as demonstrated via the experimental measurements. The formation with higher thermal maturity has lower water-wettability which could increase the brine permeability. Results demonstrated that water production is significantly affected by thermal maturity of organic matter in organic-rich mudrocks. The outcomes of this paper can potentially contribute to a better understanding of the parameters affecting water production in organic-rich mudrocks.
The new-generation oil-base mud (OBM) microresistivity imagers provide photorealistic high-resolution quantified formation imaging. One of the existing interpretation methods is based on composite processing providing an apparent resistivity image largely free of the standoff effect. Another one is the inversion-based workflow, which is an alternative quantitative interpretation, providing a higher quality resistivity image, button standoff, and formation permittivities at two frequencies. In this work, a workflow based on artificial neural networks (NNs) is developed for quantitative interpretation of OBM imager data as an alternative to inversion-based workflow.
The machine learning approach aims to achieve at least the inversion-level quality in formation resistivity, permittivity, and standoff images an order of magnitude faster, making it suitable for implementation on automated interpretation services as well as integration with other machine learning based algorithms. The major challenge is the underdetermined problem since OBM imager provides only four measurements per button, and eight model parameters related to formation, mud properties, and standoff need to be predicted. The corresponding nonlinear regression problem was extensively studied to determine tool sensitivities and the combination of inputs required to predict each unknown parameter most accurately and robustly. This study led to the design of cascaded feed-forward neural networks, where one or more model parameters are predicted at each stage and then passed on to following steps in the workflow as inputs until all unknowns are accurately obtained.
Both inverted field data sets and synthetic data from finite-element electromagnetic modeling were used in multiple training scenarios. In the first strategy, field data from few buttons and existing inversion results were used to train a single NN to reproduce standoff and resistivity images for all other buttons. Although the generated images are comparable to images coming from inversion, the method is dependent on the availability of field data for variable mud properties, which at the moment limits the generalization of the NNs to diverse mud and formation properties.
In the second strategy, we utilized the synthetic responses from a finite element model (FEM) simulator for a wide range of standoffs, formation, and mud properties to develop a cascaded workflow, where each stage predicts one or more model parameters. Early stages of the workflow predict the mud properties from low formation resistivity data sections. NNs then feed the estimated mud angle and permittivities at two frequencies into next stages of the workflow to finally predict standoff, formation resistivity, and formation permittivities. Knowledge of measurement sensitivities was critical to design the efficient parameterization and robust cascaded neural networks not only due mathematically underdetermined nature of the problem but also the wide dynamic range of mud and formation properties variation and the measurements. Results for processed resistivity, standoff, and permittivity images are presented, demonstrating very good agreement and consistency with inversion-generated images. The combination of two strategies, training on both synthetic and field data, can lead to further improvement of robustness allowing customization of interpretation applications for specific formations, muds, or applications.
Elastic anisotropy resulting from shale lamination makes fracture prediction in shale more complex, and traditional methods to predict fracture geometry assuming isotropy frequently prove to be inadequate. Common 3D fracture-modeling software is based on isotropic rock models, and models that account for anisotropy are computationally expensive, especially when numerous simulations must be performed by varying the input parameters for parametric study.
A new workflow was created that integrates anisotropic acoustic log interpretation, 3D fracture modeling, and neural networks to improve fracture prediction accuracy and efficiency for anisotropic shales. The workflow generates a neural network with a limited number of 3D fracture-modeling cases; the fracture modeling uses rock mechanical properties interpreted from sonic logs with properly selected anisotropic acoustic models. The neural network trained from a pilot/offset well can be applied to predict fracture geometries or to optimize fracturing design for other wells from the same geological basin in a timely and cost-effective manner.
The workflow is demonstrated by generating neural-network models for two shale reservoirs. The fracture geometry predicted from the anisotropic models is compared with the one predicted from the conventional isotropic simulator. The results show that ignoring shale anisotropy leads to overestimated fracture widths and underestimated fracture containments, lengths, and net-pressures. The neural-network models are run in large parametric studies to demonstrate how the effective propped length and fracture productivity varies with perforation position, injected volume, and pumping rate in the two shale formations. The results provide valuable insights of selecting perforation location and optimizing pumping strategy.
The combination of hydraulic fracturing and horizontal drilling has made production from shale and tight formations commercially realistic. However, because of the laminated and platy nature intrinsic to shales, the isotropic acoustic model, which computes a single Young’s modulus and a single Poisson’s ratio from sonic and density logs, cannot fully describe their elastic behavior.