Search

We propose a novel approach based on Deep Learning methods using a convolutional neural networks (CNNs) to compare observed four-dimensional (4-D) seismic data with reservoir simulation model results and select the simulation models with the best match. Our approach advantages are twofold: 1) it is not a pixel-based approach, as it learns spatial features through training examples and 2) it does not require the simulation model and 4-D seismic maps to be in the same domain, enabling a direct comparison of maps in the seismic domain with maps in the numerical simulation domain. Additionally, we present a new methodology to quantitatively assess the performance of each method relative to the expected results. For this purpose, we build an annotated dataset containing the maps that best match the 4-D seismic in different realistic scenarios while taking into account the specialists' knowledge. We apply our methodology to evaluate if a method is able to perform according to what specialists expect, considering the annotated data. We also propose an improvement to traditional methods using a statistical method to convert maps of different physical properties into a common property, which can be used as an alternative to avoid forward and/or inversion procedures commonly used to bring 4-D seismic and simulation data to the same domain. Finally, we present an extensive comparison, qualitative and quantitative, of our approaches with other methods in the literature, such as the traditional pixelwise strategy for maps in the same domain, and methods based on binarization of maps in different domains. The results show that our proposed methods are capable of more accurately identifying the best-matched models, according to the specialists' answers.

History matching (HM) is an inverse problem where uncertainties in attributes are reduced by comparison with observed dynamic data. Typically, normalized misfit summarizes dissimilarities between observed and simulation data. Especially for long-time series, objective functions (OFs) aggregate multiple events and tendencies relevant to field performance in a single indicator (e.g. water rate and breakthrough time). To capture the attributes influencing the reservoir behavior, we evaluate the assimilation of data series through additional OFs, obtained from splitting time-series data. In this study, two additional OF groups supplement the time-series misfits: Breakthrough Deviation (BD) indicating dissimilarities in water breakthrough time; Productivity Deviation (PD), representing mismatches of the well potential, mainly impacting the transition from history to forecast conditions. The Productivity Deviation (PD) is adapted from previous studies. Instead of simulating the last time of the historical period under forecast conditions, we propose keeping it under historical data. The change is the historical data used as target condition to the simulator: Bottom Hole Pressure (BHP) in place of liquid production and water injection rates; with this, we estimate a mismatch in well productivity, while avoiding the influence of other boundary conditions in the evaluation. Two applications (1 & 2), assimilating different OF quantities, highlight the influence of the additional groups. Application 1 only computes time-series misfit (64 OFs) whereas Application 2 includes the BD and PD (counting 128 OFs). The iterative HM method presents flexibility regarding OFs assimilated and incorporation of uncertain attributes. UNISIM-I-H case allows us to evaluate the HM considering history and forecast data. We examine differences between the 450 scenarios resulting of data assimilation for each application through four perspectives. Application 2 resulted in scenarios with better predictability of the field behavior and smoother transitions between field history and forecast periods. Field cumulative oil production of Application 2 is also forecasted closer to the reference data when compared to Application 1; all forecast periods (1, 5 and 19 years) emphasize this impact. Some wells presented breakthrough time closer to the reference for Application 2. The challenging achievement of exact BD matches leads to the third advantage of the additional indicators. These OFs supply supplementary information to the diagnosis of scenarios, identifying unnoticed problems in the traditional approach. Finally, even with an overall better performance, some of the well OFs presented poorer matches for Application 2. To explain this, we analyzed the relationship between attributes and the OFs used to update the attributes. In conclusion, the improved forecast of the simulation scenarios indicates that superior performance of the HM process is possible by splitting the available dynamic data in relevant additional OFs. This study presents a case application with 11 years of field history, in which additional OFs, derived from dynamic data, add value to the reservoir characterization. They allow the influence of uncertain attributes to be captured for relevant events in reservoir performance. We also show how the increased quantity of OFs assimilated makes the HM harder for some OFs.

The current work introduces an automatized process coupled with geological modeling to reduce uncertainty in reservoir properties. The method assimilates dynamic data of wells interactively, and applies the deviations to constrain uncertainties. The methodology deals with scalar (e.g. rock compressibility) and spatial (e.g. porosity) attributes at same time, employing specialized uncertainty reduction procedures. The procedure reduces the uncertainty of scalar attributes through Iterative Discrete Latin HyperCube method (IDLHC). To reduce uncertainties of spatial attributes, we worked on an extension of a regionalized co-simulation (co-DSS) method. The main contributions regard the proposition of update, at same time, both kinds of uncertainties (spatial and scalar); the definition of sequential rules, that simplify the process execution and avoiding subjectivities on the coupling of the geological modeling on data assimilation, as well as the automation of the process. The procedure was validated under a siliciclastic black-oil benchmark field (UNISIM-I-M), established based on Namorado Field, Campos Basin, Brazil. The procedure reduced the range of uncertainty of the scalar attributes, centralizing final PDFs with the values presented in the reference model, without collapse to a particular level, as well as preserved the geological consistency throughout data assimilation, obtaining porosity responses in agreement with the reference porosity distribution. The potential of the procedure is supported by the consistent production forecast observed in the outcomes.

Orientadores: Denis José Schiozer, Alessandra Davolio Gomes Dissertação (mestrado) - Universidade Estadual de Campinas, Faculdade de Engenharia Mecânica e Instituto de Geociências Resumo: A modelagem numérica de reservatórios petrolíferos envolve muitas incertezas, as quais podem resultar na construção de modelos com previsões de produção distintas das observadas na unidade de produção fazendo com que estes não representem adequadamente a dinâmica de produção do reservatório. Na indústria do petróleo, o procedimento aplicado para reduzir os desvios observados e consequentemente obter modelos que representem adequadamente a dinâmica de produção é conhecido como ajuste de histórico. Devido à relevância do processo de ajuste de histórico, este tem sido exaustivamente estudado e aperfeiçoado, mas ainda apresenta diversos desafios. Com o objetivo de aumentar a robustez do processo, existem três abordagens que tem despertado interesse. A primeira está ligada a aplicação de processos probabilísticos, que possibilitam uma melhor avaliação das incertezas, sendo uma abordagem mais consistente do que a abordagem determinística tradicional. A segunda se refere ao uso de múltiplas funções-objetivo locais para mensurar a qualidade do ajuste; trabalhos recentes demonstram que esta abordagem é mais indicada do que o uso da função-objetivo global tradicional. A terceira refere-se ao uso de outras fontes de dados, como a sísmica 4D, para o melhor condicionamento dos modelos numéricos de simulação. Neste contexto, Maschio et al. (2015) formalizaram uma metodologia que visa incorporar todos estes aspectos, porém com passos ainda subjetivos. Com isso, o presente trabalho propõe uma extensão a esta metodologia com o propósito de sistematizar alguns passos do processo. A principal inovação apresentada está ligada ao processo de redução de incertezas proposto, o qual consiste na atualização dos atributos incertos baseado nos modelos de simulação que apresentaram os melhores ajustes, através de um processo sistemático. Baseado em um reservatório de média complexidade, a metodologia foi validada em duas fases de aplicação. Na Fase I, foi desenvolvido o processo de calibração e redução de incertezas, utilizando somente dados de poço como fonte de dados para o condicionamento dos múltiplos modelos de simulação, enquanto que na Fase II, mapas de variação de pressão (?p) e de saturação de água (?Sw), oriundos da sísmica 4D, foram adicionados de forma quantitativa ao processo. Os resultados obtidos ilustraram a eficácia do processo proposto, o qual possibilitou uma redução acentuada das incertezas obtendo múltiplos modelos de simulação ajustados em um pequeno número de iterações Abstract: The numerical reservoir modelling is surrounded by uncertainties. These can lead to constructing numerical models which present production forecast different from observed production. In the petroleum industry, the procedure applied to reduce the observed differences, and consequently obtain adequate models, is known as history matching. Although this process has been exhaustively studied and improved, diverse challenges still exist. Aiming to improve the process robustness, three approaches have aroused: (1) application of probabilistic procedures that enables to better evaluate uncertainties, a more consistent approach than the traditional deterministic one; (2) the evaluation of multiple local objective functions to measure the quality of the match, more suitable than the traditional global objective function; (3) the additional incorporation of observed data as, for instance 4D seismic, to improve the conditioning of the numeric simulation models. In this context, Maschio et al. (2015) formalized a general methodology that aims to incorporate all these aspects, however, with subjective steps. The current work proposes an extension to this methodology with the target to systematize some of these steps. The main innovation is in the proposed uncertainty reduction process, which consists in updating the uncertain attributes based on the simulation models that presented the best matches. The validation of the proposed methodology is presented in two phases of application in a reservoir of medium complexity. In Phase I, the procedure of calibration and uncertain reduction is developed using only well data to condition the multiple models; while in Phase II, the maps of pressure changes (?p) and water saturation changes (?Sw) from 4D seismic data are integrated in a quantitative way as additional constraints. The obtained results showed the efficiency of the proposed methodology, which provided a clear reduction in the uncertainties, obtaining multiple matched models in less iterations Mestrado Reservatórios e Gestão Mestre em Ciências e Engenharia de Petróleo Brasil/MS