Naturally fractured reservoirs (NFRs) hold a significant amount of the world's hydrocarbon reserves. Compared to conventional reservoirs, NFRs exhibit a higher degree of heterogeneity and complexity created by fractures. The importance of fractures in production of oil and gas is not limited to naturally fractured reservoirs. The economic exploitation of unconventional reservoirs, which is increasingly a major source of short- and long-term energy in the United States, hinges in part on effective stimulation of low-permeability rock through multi-stage hydraulic fracturing of horizontal wells. Accurate modeling and simulation of fractured media is still challenging owing to permeability anisotropies and contrasts. Non-physical abstractions inherent in conventional dual porosity and dual permeability models make these methods inadequate for solving different fluid-flow problems in fractured reservoirs. Also, recent approaches for discrete fracture modeling may require large computational times and hence the oil industry has not widely used such approaches, even though they give more accurate representations of fractured reservoirs than dual continuum models. We developed an embedded discrete fracture model (EDFM) for an in-house fully-implicit compositional reservoir simulator. EDFM borrows the dual-medium concept from conventional dual continuum models and also incorporates the effect of each fracture explicitly. In contrast to dual continuum models, fractures have arbitrary orientations and can be oblique or vertical, honoring the complexity and heterogeneity of a typical fractured reservoir. EDFM employs a structured grid to remediate challenges associated with unstructured gridding required for other discrete fracture models. Also, the EDFM approach can be easily incorporated in existing finite difference reservoir simulators. The accuracy of the EDFM approach was confirmed by comparing the results with analytical solutions and fine-grid, explicit-fracture simulations. Comparison of our results using the EDFM approach with fine-grid simulations showed that accurate results can be achieved using moderate grid refinements. This was further verified in a mesh sensitivity study that the EDFM approach with moderate grid refinement can obtain a converged solution. Hence, EDFM offers a computationally-efficient approach for simulating fluid flow in NFRs. Furthermore, several case studies presented in this study demonstrate the applicability, robustness, and efficiency of the EDFM approach for modeling fluid flow in fractured porous media. Another advantage of EDFM is its extensibility for various applications by incorporating different physics in the model. In order to examine the effect of pressure-dependent fracture properties on production, we incorporated the dynamic behavior of fractures into EDFM by employing empirical fracture deformation models. Our simulations showed that fracture deformation, caused by effective stress changes, substantially affects pressure depletion and hydrocarbon recovery. Based on the examples presented in this study, implementation of fracture geomechanical effects in EDFM did not degrade the computational performance of EDFM. Many unconventional reservoirs comprise well-developed natural fracture networks with multiple orientations and complex hydraulic fracture patterns suggested by microseismic data. We developed a coupled dual continuum and discrete fracture model to efficiently simulate production from these reservoirs. Large-scale hydraulic fractures were modeled explicitly using the EDFM approach and numerous small-scale natural fractures were modeled using a dual continuum approach. The transport parameters for dual continuum modeling of numerous natural fractures were derived by upscaling the EDFM equations. Comparison of the results using the coupled model with that of using the EDFM approach to represent all natural and hydraulic fractures explicitly showed that reasonably accurate results can be obtained at much lower computational cost by using the coupled approach with moderate grid refinements. / text

[pt] Nesta dissertação é apresentado um estudo comparativo entre dois métodos de descrição do escoamento de fluidos em meios porosos heterogêneos e compostos de elementos arquiteturais de permoporosidade elevada, chamados fraturas e carstes. Fraturas são descontinuidades nas formações desencadeadas por tensão mecânica e carstes são espaços vazios na rocha formados por dissolução. Ambos são característicos das rochas carbonáticas do Pré-Sal, principais reservatórios produtores de petróleo do Brasil. A presença destes elementos traz grande dificuldade de caracterização e, por consequência, introduz grande incerteza nas curvas de produção previstas para cada campo. Os simuladores numéricos em diferenças finitas de Engenharia de Reservatórios, responsáveis pela geração destas curvas, representam de forma aproximada o escoamento nos espaços vazios da rocha devido à formulação aplicada. Além disso, devido à escala de quilômetros de extensão em que normalmente são utilizados, usam células que incluem os três meios (matriz porosa, carste e fratura), cujas escalas são menores que a de cada célula. Nesta dissertação, busca-se entender a equivalência entre simuladores numéricos black oil e o modelo de Brinkman, ainda não utilizado amplamente, que é um modelo físico constituído por equações que representam mais fielmente o escoamento, tanto na matriz porosa, quanto nas regiões de altíssima porosidade e nas fronteiras entre elas. Para esse objetivo, foi desenvolvido um simulador de Brinkman de fluxo monofásico em duas dimensões e capaz de representar o regime não permanente, utilizando o método dos elementos finitos para resolução das equações diferenciais. Os comportamentos não permanente e permanente do simulador criado foram validados por meio de exemplos da literatura. Os valores obtidos para a propagação da pressão e velocidade de fluxo foram comparados com os de um simulador numérico black oil na reprodução do escoamento em camadas selecionadas do modelo do carbonato do Lajedo Arapuá. Foram realizados estudos com modelos de duplo contínuo (que representam, através de parâmetros específicos, o fluxo na matriz porosa e nas fraturas, com um termo de transferência entre eles) buscando, através da variação da permeabilidade efetiva da fratura e do fator de forma, convergência aos resultados do modelo de contínuo único de referência, selecionado após análise dos resultados comparativos com Brinkman. Os resultados mostraram pouca variação entre os resultados dos métodos em cenários nos quais o sistema cárstico é composto por vugs dispersos e desconectados, enquanto que carstes em forma de condutos e com configurações complexas causam alterações na propagação da onda de pressão e nas velocidades de fluxo entre os modelos, principalmente em cenários com valores de permeabilidade mais próximos entre carste e matriz porosa na simulação black oil. A análise em duplo contínuo mostrou ser possível obter, através de modelos homogêneos e com escala até 10 vezes maiores, resultados semelhantes aos obtidos com uma modelagem black oil heterogênea com caracterização do sistema cárstico. Também foi possível concluir que a permeabilidade efetiva de fratura é suficiente como parâmetro de ajuste para encontrar um modelo equivalente, dentro de um critério de valor de variação, ao modelo de contínuo único. / [en] This dissertation presents a comparative study between two porous media fluid flow description methods applied to heterogeneous reservoirs composed of architectural elements of high permo-porosity, called fractures andkarsts. Fractures are formation discontinuities triggered by mechanical stressand karsts are empty cavities inside the formation generated by dissolution.Both are characteristic of Pre-Salt carbonate rocks, which are the main oil producing reservoirs in Brazil. The existence of these elements brings complexityin characterization and, hence, increases the uncertainty in field productionprediction curves. The Reservoir Engineering numerical simulators, based onfinite differences, that generate these curves model the flow behavior insidethe cavities in an approximate manner due to the mathematical formulationapplied. Besides, due to the kilometric scale in which they are usually applied,these models use cells that comprehend the three media (matrix, karst andfractures), whose scales are smaller than the cell’s. This dissertation evaluatesthe equivalence between black oil simulation and the Brinkman model, still notwidely used, which is a physical model made of equations that represent porousmedia flow inside high porosity regions as well as at the porous matrix andits boundaries. With this objective, a two-dimensional single-phase Brinkmansimulator, capable of representing transient flow, was designed using the finiteelements method to solve differential equations. The simulator s transient andpermanent behaviors were validated through literature analytical solutions.The pressure propagation and flow velocity values obtained while simulatingfluid flow inside selected layers from Lajedo Arapuá s carbonate formation werecompared to the ones from a black oil simulator. Studies were performed usingdual-continuum models (which calculate the fluid flow inside porous matrixand fractures separately and use a transfer term to account for the flow between them) seeking, through variation of the fracture effective permeabilityand the shape factor, convergence to the mono-continuum reference model thatwas selected after Brinkman s results comparative analysis. The results showsmall variation between the two methods when the karst system is composedof sparse and disconnected vugs, while conduit shaped karsts with complexconfiguration increase the variation in pressure wave propagation and flowvelocity values between models, especially in scenarios where matrix permeability values were closer to karst permeability values in the black oil model.The dual-continuum analysis showed that it is possible to obtain, through homogeneous and even ten times coarser models, similar results to those obtainedby a heterogeneous black oil model with karst system characterization. It wasalso possible to conclude that fracture effective permeability was sufficient asa fitting parameter in order to achieve equivalent results to those from themono-continuum model, using a threshold criteria.