|Theme||Visible||Selectable||Appearance||Zoom Range (now: 0)|
We present an integrated transfer learning approach that enables the reuse of modeling and optimization knowledge accumulated from previously learned development scenarios to accelerate the execution of forthcoming development scenarios. Transfer of this knowledge can be applied between different wells, pads or fields. Therefore, the main objective of the proposed methodology is to create efficient and scalable field development workflows involving many well/pad configurations, decision and uncertainty parameters.
Data, simulation, forecasting and optimization requirements can vary significantly according to the number of wells in operation and to be incorporated in each field plan. When the number of wells is relatively small (say a few dozens) and detailed data is available, full-physics simulations can be afforded for further increasing the reliability of the analysis process. When planning involves a few hundreds of wells at the level of several pads or the field, data usually becomes less detailed and analytical models based on production observations become more suitable to speed up the analysis. As we incorporate thousands of wells and/or move towards the extreme of a basin, the data becomes much coarser and scattered. Moreover, a large number of wells makes impractical the use of analytical models through a manual process. In this case, the field analysis is usually replaced by statistical or machine learning approaches that provide a holistic view of geo-spatial properties against production trends. This is usually achieved by a means of cross-field analogies and data aggregation methods (e.g., clustering and type curve analysis).
Figure 1 shows the modeling spectrum subject to the situation just described. At any intermediate point of this spectrum there are usually a few models or tools that can handle different types of physics and scales. In many cases, selecting the right model or workflow is a challenge and, especially when they have to adapt to different future data requirements, scales and decisions. There is even a major overpressing issue: outcomes of these models or decisions are discarded once field problem specifications change. For example, large rounds of simulations are repeated when the permeability field, the number of decision parameters or range uncertainties are revisited.
Hence, despite how much data is available or how adequate are models to deal with different scenarios, computational knowledge accumulated during the life of an asset is usually disregarded. Besides, gained human experience, traditional modeling and optimization tools are unable to leverage this knowledge to speed up forthcoming decision cycles. Therefore, the efficiency and predictability range of these tools is highly dependent on existing domain expertise to be able to bridge geological, operational and decision parameters with analogous field scenarios.
Tang, Hewei (Texas A&M University) | Yan, Bicheng (Texas A&M University (now with Sanchez Oil and Gas)) | Chai, Zhi (Texas A&M University) | Zuo, Lihua (Texas A&M University) | Killough, John (Texas A&M University) | Sun, Zhuang (University of Texas at Austin)
Well interference is a common phenomenon in unconventional-reservoir development. The completion and production of infill wells can lead to either positive or negative well-interference impacts on the existing producers. Many researchers have investigated the well-interference phenomenon; however, few of them attempted to apply rigorous simulation methods to analyze both positive and negative well-interference effects, especially in two different formations. In this work, we develop a comprehensive compositional reservoir model to study the well-interference phenomena in the Eagle Ford Shale/Austin Chalk production system. The reservoir model considers capillary pressure in the vapor/liquid-equilibrium (VLE) equation (nanopore-confinement effect), and applies the embedded discrete-fracture model (EDFM) for dynamic fracture modeling. We also include a multisegment-well model to characterize the wellbore-crossflow effect introduced by fracture hits. The simulation results indicate that negative well-interference impact is much more common in the production system. With a smaller permeability difference, the hydraulic-fracturing effect can lead to a positive well-interference period of several hundred days. The nanopore-confinement effect in the Eagle Ford Shale can contribute to the negative well-interference effect. We also analyze the impact of other factors such as initial reservoir pressure, matrix porosity, initial water saturation, and the natural-fracture system on the well performance. Our work provides valuable insights into dynamic well performance under the impact of well interference.
The fracture description plays an important role in shale gas well production performance prediction, late production refracturing design and infill well trajectory design. Based on the development and geological parameters of Fuling shale gas field, the enhanced discrete fracture network (EDFN) numerical simulator is used to study the influence of fracture length, total fracture length, stage spacing and relative position of fractures on the contribution ratio of fracture stage. According to the relationship among JY46-3HF gas production profile, gas production contribution ratio and fracture characteristic parameters, a fracture network model is established. The simulation results of gas production contribution ratio of each fracture stage are highly consistent with the measured data. The research results show that: the contribution ratio of gas production in each fracture stage is positively related to the total cumulative fracture length, and the fracture spacing and relative position of fractures affect the contribution ratio of fracture stage to shale gas well by the size of matrix area controlled by fracture.
Well interference has become a major concern in current shale reservoir development. Inter-well fracturing interference raised by drilling infill well/well pad in shale reservoir can lead to either beneficial or detrimental impact on existing producer. To better simulate the wellbore cross flow raised by fracture communication and the well performance under well group control, we fully couple a generalized multi-segment well (MSW) model to our in house 3D compositional simulator. The model discretizes wellbore into segments, allowing a comprehensive representation of fluid flow dynamic in the wellbore. The changing of hydraulic fracture network due to infill wells is modeled by Embedded Discrete Fracture Modeling (EDFM), which avoids the complication of regenerating the reservoir mesh.
We first create two extreme cases to illustrate possible mechanisms for the positive and negative well interference effects. Furthermore, we introduce a 3D case with parent well completed vertically above a two-well infill pad. The joint effect of wellbore cross flow and stimulated reservoir volume (SRV) overlap can be observed in this case. We further investigate the impact of interference between two infill wells on the cumulative production of parent well and infill well pad. The simulation results demonstrate the capability of the coupled MSW-reservoir model in simulating different well interference scenarios. The model can be a flexible tool in analyzing and forecasting the interfered well production performance.
The development of unconventional shale reservoirs requires sufficient drainage of resources through drilling multiple well pads and multi-stage hydraulic fracturing. The resulting complex fracture networks improve the recovery from these reservoirs by dramatically increasing the well-reservoir contact area (Xue et al. 2016). However, the same extensive fracture network may also increase the risk of “fracture hits” as well spacing decreases (Ajani and Kelkar 2012; Yu et al. 2016).
Alfi, Masoud (Texas A&M University) | Chai, Zhi (Texas A&M University) | Yan, Bicheng (Texas A&M University) | Stimpson, Brian C. (Texas A&M University) | Barrufet, Maria A. (Texas A&M University) | Killough, John (Texas A&M University)
This paper tackles several challenging aspects of shale reservoir simulation including the complex physics of fluid transportation and the corresponding reservoir simulator development, fractured media modeling, and modified PVT behavior in nano-pores. In the first part of the paper, a finite-volume based compositional simulator, named GURU, is developed to incorporate the study of more advanced physics in shale reservoirs. Our fit-for-purpose simulator is able to provide a multi-porosity/multi-permeability simulation capability, which is a top priority in simulation of highly heterogeneous shale resources. The performance of our simulator in hydraulically fractured reservoirs is evaluated through some case studies. An Embedded Discrete Fracture Model (EDFM) is discussed in the second part of paper to handle arbitrary fracture geometry flexibly. Our EDFM model inherits the advantages of using a multi-porosity model to replace laborious traditional approaches to characterize multi-scale reservoir heterogeneity and the complex fracture network in shale reservoirs. In the last part of this paper, we have provided an in-depth discussion on how rock-fluid interactions can affect the phase behavior and fluid distribution in shale. Capillary pressure effect on phase behavior of reservoir fluid is analyzed where a dynamic interfacial tension calculation process is incorporated to account for the effect of pore size and fluid composition in capillary pressure calculations. In addition to that, application of a modified Peng-Robinson equation of state shows that the tight porous media in shale can act as a semi-permeable membrane and selectively filters molecules based on their sizes. This phenomenon is further investigated in a so-called ‘sieving or size filtration effect’.
Low-permeability shale formations have received a great deal of attention because of their potential to supply the entire world with sufficient energy for the decades to come. During the past few years, there has been a substantial effort to understand the behavior of such reservoirs. Despite the tremendous effort, it appears that studies, understandings, and effective technologies required for development of such resources are still behind the industry needs. Part of this problem is attributed to complexities, such as geological heterogeneity, fluid transport mechanisms, and fluid properties in small pores of low-permeability shale reservoirs.
The coupled flow-geomechanics model is required to investigate the stress change, rock-compaction behavior, and stress-dependent properties in many reservoir scenarios. However, the coupled model for large-scale or three-dimensional simulation problems usually encounters large matrix system and high computation expenses, where the time stepping is a crucial factor for numerical stability and computational efficiency. In this paper, an adaptive time stepping with the modified local error method was presented to reduce iteration time and improve the computation efficiency for the coupled flow-geomechanics modeling. Firstly, the iterative coupling approach with the fixed-stress method was introduced, where the flow and geomechanics equations are sequentially solved at each time step. Secondly, due to updating geomechanics module consumes the major computing time for the coupled problems, the modified local error method was mainly used for geomechanics module, where fewer geomechanics time steps are needed after implementation. Specifically, the geomechanics module will be updated until a given local error of displacement is reached, and the time step size will be automatically adapted based on the change of displacement, which is more efficient than the constant time step method. Finally, a synthetic two-dimensional coupled production problem is established to apply the proposed adaptive time stepping approach, where the numerical results including the computing efficiency are compared with the results from regular sequential method and the fully coupled model. The sensitivity about the local error tolerance was also investigated.
The geomechanical responses regarding mechanical loading is validated by comparing with the analytical solution of Terzaghi's consolidation problem. The numerical results about the pressure and displacement change on the two-dimensional coupled model are compared with the results from the regular sequential method and the fully-coupled method. The modified local error method, which adjusts time step size for both flow and geomechanics module, not only yields a higher-order solution for better accuracy, but also significantly reduces the iteration number and computation time, especially for the cases with low truncation error requirement. The error tolerance of displacement is critical on when the step size will be adapted. Small error tolerance can maintain the accuracy while it needs more iteration computing. The strategy about how to modify step size plays an important role in the stability and computing efficiency for the modified local error method. A large increase or cut on one-step size could bring the oscillation results. Overall, the adaptive time stepping approach can both effectively reduce total computation time and simultaneously maintain the accuracy for the coupled flow-geomechanics problems. It is useful for large-scale or three-dimensional coupled problems, where the high computational efficiency is required.
In this paper, geomechanics is coupled with reservoir flow for modeling the depletion and deformation in fractured vuggy carbonate reservoir. Different from the dual- and triple-porosity models or the coupled approaches in which the vugs are considered as a continuous porosity, the vugs are treated as virtual volumes in this study. For each vug, the fluid exchange at the vug-matrix interface is dynamically calculated with time evolution and the pore pressure in the vugs is updated through considering both the fluid material balance and the volume change due to the mechanical deformation of vug. The fluid-mechanical interaction in the rock matrix and natural fractures is calculated based on the framework of Biot's poroelstic theory. The mechanical and hydraulic interactions between vugs and matrix are preserved and the stress evolution due to the depletion can be dynamically updated. The results in this study show that, the depletion process is mainly controlled by the fluid storage of the vugs. Fluid modulus is thus a more sensitive parameter than the rock/fracture modulus in terms of the depletion. However, the rock/fracture modulus can also affect the deformation of the system and thus affect the volume and pressure changes of the vugs.
An, Cheng (Texas A&M University) | Fang, Yi (Pennsylvania University) | Liu, Shuangshuang (PetroChina) | Alfi, Masoud (Texas A&M University) | Yan, Bicheng (Texas A&M University) | Wang, Yuhe (Texas A&M University) | Killough, John (Texas A&M University)
During reservoir depletion, effective stress is increased and permeability is reduced, while the organic- rich matrix might experience a shrinkage process that will boost the permeability. The main objective was to develop a mathematical simulator coupling gas flow process, geomechanics effects, and matrix shrinkage in order to evaluate their influences on reservoir permeability and production performance. The mesh was divided into three different continuums: organic matter, inorganic matter, and natural fractures. Matrix shrinkage was only considered for organic matter because of gas desorption, and the stress- dependent permeability was considered for both inorganic matter and natural fractures. The flow and stress-equilibrium equations were solved by the fixed-stress sequential method, where the flow equations are solved first, followed by the mechanics equations. The displacements are solved for each grid node by finite element method, and the pressure is solved by the integral finite difference method. Different stress- dependent correlations are chosen to separately apply to the three porous media. Based on those correlations, the porosity and permeability are updated at end of each time step. A synthetic reservoir model was built, where the permeability change and the accumulative gas production is calculated at each time step. The results of permeability change and gas production rate are compared for three different cases: the coupled flow and geomechanics model without permeability change, the coupled model considering stress-dependent permeability, the coupled model considering both matrix shrinkage and permeability change. Additionally, the sensitivity analyses were investigated for total organic carbon (TOC), Young's modulus, matrix permeability, and bottom hole pressure.
Results show that the stress-dependent permeability plays a large influence on the gas production performance, because permeability could be significantly reduced with the decrease of reservoir pressure. The matrix shrinkage on organic matter could provide an obvious rebound on accumulative production at the late producing stage, because the permeability is boosted by that media shrinkage at the late producing stage. That explains why permeability largely decreases at early stage and then gradually reduces in experimental data. However, the permeability and production loss are highly depended on the selected correlation, its coefficients, reservoir initial condition, and rock properties. Organic matter is the critical controller on matrix shrinkage: the higher the TOC, the larger the increase of permeability. Nevertheless, their overall impacts on production is quite limited. Young's modulus does not make obvious differences on the accumulative gas based on the numerical results. The large matrix permeability and higher bottom hole pressure can reduce the production loss caused by the effect of stress-dependent permeability. Overall, the triple-porosity coupled simulator can quantitatively interpret the impacts of matrix shrinkage and geomechanics effect on permeability and gas production performance for organic-rich shale reservoirs. This provides more realistic production performance evaluation and economic assessment when the stress-dependent permeability needs to be considered.
The computation mesh is one of the important challenges for the simulation of the discrete fracture networks (DFN). In this work, an efficient mesh method is introduced to tackle this problem.
First, convert the map with complexity fracture network into a numerical image comprised
The results show that this method is appropriate to be applied in reservoirs with different geometry and arbitrary complex fracture networks.
Alfi, Masoud (Texas A&M University) | An, Cheng (Texas A&M University) | Cao, Yang (Texas A&M University) | Yan, Bicheng (Texas A&M University) | Barrufet, Maria A. (Texas A&M University) | Killough, John E. (Texas A&M University)
We propose a robust multiple permeability model to account for pore size variability and filtration effects in shale reservoirs and to explain the behavior of unconventional resources such as anomalous gas oil ratios. Recent developments in technology transformed unconventional low-permeability shales into reliable energy sources. With regard to capacity, unconventional resources are found to be more abundant than conventional ones. These resources have opened new challenges and technical difficulties along with enhanced capacities. Recent advances in high precision analytical tools have revealed that pore size distribution in shale reservoirs cover a wide range. Molecules in pores with different sizes may exhibit significantly different thermodynamic behavior. Rock fluid interactions and space hindrance effects play an important role when pore sizes become close to species' molecular dimeters. This effect can result in a composition difference between pores with large and small diameters in shale reservoirs (sieving effect), with small pores mostly filled with smaller hydrocarbon molecules and large molecules residing in larger pores. To account for such a diverse behavior, this paper proposes a multiple permeability model, which divides shale media into three different permeability/porosity systems: fracture, matrix with large pores and matrix with small pores. We use a modified version of the Peng Robinson equation of state to model the equilibrium hydrocarbon distribution in large and small pores. Our thermodynamics calculations show that as pore dimeter decreases, the concentration of larger hydrocarbon molecules in those pores decreases because of size filtration. A synthetic reservoir model is used along with the multiple permeability model to analyze reservoir production behavior at different conditions. The so-called sieving effect is believed to be responsible for the anomalous production behavior (lower-than-expected or constant gas oil ratios for extended production periods). Our model is then applied to a real Eagle Ford case to history-match production data with the stimulated reservoir volume built upon the fracture microseismic data. The results show that our multiple permeability model provides a powerful tool to evaluate the complicated flow dynamics in liquid shales.