Scientific Computing

A dynamic multilevel method for fully-coupled simulation of flow and heat transfer in heterogeneous and fractured geothermal reservoirs is presented (FG-ADM). The FG-ADM develops an advanced simulation method which maintains its efficiency when scaled up to field-scale applications, at the same time, it remains accurate in presence of complex fluid physics and heterogeneous rock properties. The embedded discrete fracture model is employed to accurately represent fractures without the necessity of unstructured complex grids. On the fine-scale system, FG-ADM introduces a multi-resolution nested dynamic grid, based on the dynamic time-dependent solution of the heat and mass transport equations. The fully-coupled implicit simulation strategy, in addition to the multilevel multiscale framework, makes FG-ADM to be stable and efficient in presence of strong flow-heat coupling terms. Furthermore, its finite-volume formulation preserves local conservation for both mass and heat fluxes. Multi-level local basis functions for pressure and temperature are introduced, in order to accurately represent the heterogeneous fractured rocks. These basis functions are constructed at the beginning of the simulation, and are reused during the entire dynamic time-dependent simulation. For several heterogeneous test cases with complex fracture networks we show that, by employing only a fraction of the fine-scale grid cells, FG-ADM can accurately represent the complex flow-heat solutions in the fractured subsurface formations.

The vast majority of grids for reservoir modeling and simulation workflows are based on pillar gridding or stairstep grid technologies. The grids are part of a feature-rich and well-established modeling workflow provided by many commercial software packages. Undesirable and significant simplifications to the gridding often arise when employing such approaches in structurally complex areas, and this will clearly lead to poor predictions from the downstream modeling.

In the classical gridding and modeling workflow, the grid is built in geological space from input horizon and fault interpretations, and the property modeling occurs in an approximated ‘depositional’ space generated from the geological space grid cells. The unstructured grids that we consider here are based on a very different workflow: a volume-based structural model is first constructed from the fault/horizon input data; a flattening (‘depositional’) mapping deforms the mesh of the structural model under mechanical and geometric constraints; the property modeling occurs in this depositional space on a regular cuboidal grid; after ‘cutting’ this grid by the geological discontinuities, the inverse depositional mapping recovers the final unstructured grid in geological space. A critical part of the depositional transformation is the improved preservation of geodetic distances and the layer-orthogonality of the grid cells.

The final grid is an accurate representation of the input structural model, and therefore the quality checking of the modeling workflow must be focused on the input data and structural model creation. We describe a variety of basic quality checking and structurally-focused tools that should be applied at this stage; these tools aim to ensure the accuracy of the depositional transformation, and consequently ensure both the quality of the generated grid and the consistent representation of the property models. A variety of quality assurance metrics applied to the depositional/geological grid geometries provide spatial measures of the ‘quality’ of the gridding and modeling workflow, and the ultimate validation of the structural quality of the input data.

Two case studies will be used to demonstrate this novel workflow for creating high-quality unstructured grids in structurally complex areas. The improved quality is validated by monitoring downstream impacts on property prediction and reservoir simulation; these improved prediction scenarios are a more accurate basis for history matching approaches.

One major challenge with the current subsurface modeling workflows is the difficulty to transfer the high complexity of static geological models to simulation. This paper describes an improved gridding technology that overcomes the shortcomings of existing corner-point stair-step and pillar-based grids in capturing the complex geological features of hydrocarbon reservoirs. It also highlights the seamless use of this new gridding technology in flow simulation.

The reservoir's geologic features are initially interpreted as 3D surfaces that are later connected to form structural models. The volume between these surfaces is then discretized into a grid object in order to represent the petrophysical, fluid, and flow properties. This paper introduces a new orthogonal, semi-structured gridding algorithm that uses a truncated cell approach to precisely capture the geometry of faults and unconformities.

After reviewing the different types of grids commonly used in geological modeling, the benefits of the new approach will be detailed, while highlighting its compatibility with commercial flow simulators.

Common modeling practices that use corner-point stair-step and pillar-based grids fail to preserve the geometry of most geological objects. This is especially true in highly faulted and erosional environments. The new gridding algorithm presented in this paper addresses three major shortcomings of the current approaches by providing an efficient way to:

Accurately represent any type of geological structures in a 3D grid. The appeal of the technique is its simplicity. The gridding algorithm relies on three components only: a surface-based structural model, a stratigraphic model and a 3D cell resolution.

This is the first high-accuracy gridding system that is designed to be simulator agnostic. This means that the approach is open and flexible enough to be used by any commercial flow simulators, giving simulation engineers a unique opportunity to run models without the need for any explicit grid system conversion.

Is the Cloud Mature Enough for High-Performance Computing? The majority of Shell’s HPC work helps support its seismic imaging operations. The company supports 45 HPC applications, with the bulk of its processing and production workload taking place in-house. Oil and gas is in the midst of a pervasive digital transformation in which the industry is changing the way it manages assets, the way it interacts with customers, and the way it develops internal workflows. Perhaps one of the most significant impacts of this transformation, however, is the way in which companies characterize their subsurface data.

The Processing Problem: Can Computers Keep Up With Industry Demand? Big data is one of the big buzzwords in oil and gas operations today. Operators cannot get enough of it. Managing a successful venture requires the ability to extract valuable information from massive data sets and process that information in a quick and efficient manner. High-performance computing (HPC) is critical in making these things happen.

Constructing reservoir models that are consistent with geophysical and geological static data is well understood. A persisting challenge is to condition such models to the available production dynamic data through the process of history-matching. A new algorithm is utilized to identify influencing grid-block properties based on analytical sensitivity calculations. The derived sensitivities allow efficient modification of grid-block properties and saturation functions to improve the history-match. This innovative approach preserves the geological model features, because changes are done at the grid-block level and are only as small as needed to achieve a good match.

Usually, the number of parameters is so immense that engineers have to either restrict their attention to a small subset of the parameters (and likely missing crucial ones), or unnecessarily pay extremely high simulation costs. In this paper, we employ a new assisted history-matching technique that computes the parameter sensitivities analytically and for each grid-block. Here, the derivatives of the mismatch with respect to each parameter are rigorously computed based on the black-oil simulator's fluid flow equations. Hence, a single simulation run followed by a derivatives calculation session is sufficient to detect how each parameter affects the mismatch, and consequently, to decide how (or whether) to change each parameter to improve the match.

This technique was successfully applied to history-match a mature oil field in the Middle East. Two sets of parameters are modified: The permeability in 3 directions per grid block, and relative permeability curves for about 50 saturation regions. The goal was to match water-cut for individual wells.

With this analytical technique, excellent improvement in the match was achieved after only a dozen simulation runs and within a couple of days. Because the modifications are at the grid block level and minimal (only as and where needed), the technique preserves the original geological features of the model to a great extent. Eliminating the need of manual local modifications (e.g. box multipliers) is an important advantage of the method. The relative permeability curves have been tweaked successfully for numerous saturation regions using Corey model parameters. The ability to adjust many curves successfully using just a few simulation runs represents a significant advancement in the field of assisted history-matching.

ABSTRACT: Injection-induced seismicity (IIS) depends on pore pressure, in-situ stress state, and fault orientation; generally occurs in basement rock that contains fractures and faults; and moves away from the injection well as a nonlinear diffusion process. Therefore, to numerically model IIS a code should incorporate flow and geomechanics, the presence of fractures and faults, and the capability for hydraulic diffusivity to evolve with effective stress and failure history. In this work, we introduce and verify a modeling framework that allows hydraulic diffusivity to evolve as fractures open and close. Details and challenges in code development are discussed, including how the Bandis model for normal fracture deformation can be used to calculate hydraulic diffusivity as a function of effective normal stress. The discrete fracture network and matrix (DFNM) model is implemented in PFLOTRAN such that hydraulic diffusivity has different constitutive relationships for fracture and matrix grid cells. This model is applied to understand the recent IIS near Greeley, Colorado, and its results are compared to: (a) a traditional DFNM model where hydraulic diffusivity cannot evolve and (b) an equivalent porous media (EPM) model where the effect of the fractures are averaged over a large region of rock. The new DFNM model predicts critical pressure will propagate farther from an injection well. This modeling framework shows promise for applications where fracture and matrix flow are important and hydraulic diffusivity is a function of pressure, stress, and/or shear failure history.

The discrete-finite element coupling method is an effective approach to simulate the complex interactions between sea ice and offshore structures and ice-induced vibrations (IIVs) of structures. However, the small time step in the discrete element method, as the time step of the coupled method, is time-consuming. Adoption of a time multiscale strategy can solve this problem. This paper proposes a coupled discrete-finite element method based on a domain decomposition method to analyze the interactions between sea ice and a conical jacket platform. Moreover, IIVs of the platform were analyzed. The computational domain is split into several subdomains based on whether sea ice interacts with the platform. The subdomains directly impacted by sea ice use small time steps of the discrete element method. The numerical results show that the proposed time-efficient method is reliable and stable for the simulations of ice-platform interactions.

INTRODUCTION

In cold regions, the vibrations of offshore platforms induced by sea ice can be harmful for not only the routine production but also the serviceability and safety of platforms. Conical jacket platforms have been used considerably in the Bohai Sea of China. The forces induced by sea ice are the dominant environment loads acting on the platforms. Ice-induced vibrations (IIVs) of platforms have also been reported by Yue et al. (2009).

To overcome IIVs of platforms, some beneficial work including field measurements, model tests, and numerical simulations has been conducted on the interactions between sea ice and offshore platforms (Huang et al., 2013; Nord et al., 2015). Because field and scale tests are difficult and expensive, numerical simulations are usually adopted for investigating the dynamic behaviors of offshore platforms under ice loads (Hopkins, 1997; Paavilainen and Tuhkuri, 2013). Kärnä and Turunen (1989) calculated the IIVs of a narrow structure by assuming ice load to be a function of the relative displacement and relative velocity between ice and the structure. The finite element method (FEM) has also been utilized in ice load analyses in which the sea ice is approximated using the material’s nonlinear model (Sand and Fransson, 2006). However, the continuum-based FEM is limited by the inherently discrete nature of sea ice, especially in the case of floe ice.

Hybrid steam and in-situ combustion recovery processes have shown advantages over pure steam injection for recovery of oil sands resources, particularly with respect to reducing costs and lowering requirement for water and natural gas. However, it has been very challenging to predict field performance of hybrid steam and combustion processes with a reasonable degree of confidence. Usually, a combustion front has a thickness of only a few inches and high resolution grids are required to capture the steep temperature, saturation and fluid composition gradients in the vicinity of the combustion front. Using high resolution, fine grids to improve accuracy of simulation requires excessive computation time and, therefore, may be impractical for field scale modelling. It is important to have a robust simulation tool to accurately predict reservoir performance without compromising the computational efficiency.

In this work, numerical modeling of a hybrid steam and combustion recovery process was performed in a typical Athabasca oil sands reservoir. A comprehensive new reaction kinetics model derived from laboratory results was incorporated to represent the complex chemical reactions in the combustion process. The hybrid recovery process utilized oxygen enriched air co-injection after several years of SAGD operation. In the numerical model, safe limits were set on producing well temperature and oxygen content of produced fluids. The initial grid size in the numerical model was at the centimeter scale resulting in large run time, and thus, in order to improve the computational efficiency, a dynamic gridding feature was applied. Parameters for controlling the creation of a dynamic grid and subsequently reverting back to the coarse grid have been examined in order to properly trigger the dynamic gridding feature in the model. Once the optimized dynamic gridding parameters were determined, several different well configurations were investigated. Comparisons were made between SAGD and hybrid steam/combustion processes in terms of cumulative water (steam) injection, cumulative oil production, and a steam-oil ratio.

By comparing the simulation results from the fine grid model and the dynamic gridding model, it has been found that the temperature gradient is the best criterion to use for controlling dynamic gridding compared to fluid saturation and/or composition criteria. The threshold value for the temperature criterion was determined to be 35°C. The model locates the fine grids in close proximity to the combustion front where the temperature and fluid saturation gradients are the steepest and it places the coarse grid blocks elsewhere in the model. Comparisons are made between the computation time and the accuracy of the simulation and these demonstrate that dynamic grid amalgamation reduces the computation time significantly while maintaining reasonable computation accuracy of simulation. Compared with SAGD, the hybrid steam/in-situ combustion process reduced cumulative water usage (steam injection) by 20% to 27%, while the cumulative oil production remained the same.

This paper provides a workflow for modelling of hybrid steam and combustion processes. Also, it is expected that this work will provide insights for field design of these hybrid thermal recovery processes.

Michael H. Weatherl, SPE, is an engineering consultant and president of Well Integrity in Scott, Louisiana. He holds a BS degree in petroleum engineering from The University of Tulsa and has been a registered petroleum engineer in Texas since 1993. Before starting Well Integrity in August 2014, Weatherl worked as a drilling and completion team leader for Hess’ New Ventures Unit in Houston following assignments in Norway and Offshore Americas. Before Hess, he worked for 25 years for Chevron, including in a number of positions in production and drilling in Louisiana and Texas. Weatherl is a member of the JPT Editorial Committee and serves on the SPE Deepwater Drilling and Completion Conference Committee.