Fast Reservoir Characterization and Development of a Field Case Study with Real Production and 4D Seismic Data
MetadataShow full item record
The primary goal of this PhD is to provide methods for continuous and fast optimization and updating of reservoir simulation models (i.e. history matching and associated forecast) based on production data, 4D seismic data and other available data. To accomplish this goal the following strategies have been employed: Comparison and Combination: Combining the use of time-lapse seismic data and production data for history matching and parameter estimation, and then comparing and combining different methods so as to establish the best strategy to achieve an optimal solution. Smart Reservoir Modeling: Fast simulation technology with seamless modelreduction capabilities (model complexity selected semi-automatically at run-time to fit time-constraints and available computer resources). Parameter reduction techniques employed in this work include principal component analysis (PCA) and discrete cosine transform (DCT), these techniques reduce the number of variables while they retain the geological properties as much as possible. Here the general goal is to speed up the process so as to enable increased use of flow of information for decision support in operations within short time horizons. Also the use of multiple processors (distributed computing environment) has been applied to solve different time-consuming problems in this work. Optimization: History matching (HM) can be done in two main ways, the old approach (manual history matching) and the modern approach (semi-automatic history matching). Semi-automatic HM can be regarded as an optimization procedure which can be carried out in three steps: mathematical model, objective function and optimization algorithm. The mathematical model (forward model) in this project consists of a reservoir simulator and a forward seismic model. A three component three-phase black oil commercial simulator (ECLIPSE 100) is used to simulate fluid and pressure changes within the reservoir due to depletion and water injection. Forward seismic modeling software, based on rock physic formulations (Gassmann equation and Hertz-Mindlin model) and matrix propagating techniques developed at NTNU are used to provide 4D seismic amplitudes from saturation and pressure changes. Then an objective function was formed that contains both 4D seismic data and production data parts. The next step was the employment of various optimization algorithms that were tested to choose the easiest, fastest and most efficient and robust optimization algorithm in conditions which are case dependent. The thesis starts by generally reviewing advanced techniques of history matching (Chapter 2) where global optimizers have shown simplicity when it comes to timelapse seismic data integration in the HM process. As obtaining gradients from the seismic attributes has been difficult and is less well developed, this restricts the applicability of gradient-based methods when incorporating seismic data. Transform domain and parameter reduction techniques such as discrete cosine transform (DCT) and principal component analysis (PCA) are employed (Chapter 3) and then particle swarm optimization method is used together with PCA (Chapter 4). PSO global exploration avoids local optima and can deal with non-smooth functions, it is easy to implement, non-invasive with respect to simulation and can be implemented in a cluster (parallelizable). The secondary objective of this thesis has been to design a benchmark database for research and trial activities and then organise a comparative case study followed by analysis and comparison of the results from different participants. The required database should use a real field and in particular promote comparative studies of alternative methods for history matching and ultimately closed loop reservoir management. Statoil’s Norne Field in the Norwegian Sea has been in production for approximately 15 years. The field has high quality 4D seismic data, production data and well logs in addition to a reservoir model (geo model) and seismic model. This data has kindly been made available to the IO Center by the Norne license in general and Statoil in particular. Through the IO Center the data have been made available in packages for IO Center partners and the research community as a whole. The assignment emphasizes the design of a benchmark case for research and the focus on the utility value of a model with real data opens this to several research communities. The first Norne case study has been organized using data from the Esegment of the Norne field; there after an applied technology workshop was prepared in June 2011 whereby 80 delegates from 11 countries worldwide met. The workshop was organized by the cooperation between the IO Center and the Society of Petroleum Engineers (SPE), and four groups from Stanford University, Norwegian University of Science and Technology (NTNU), TU Delft and Texas A&M University presented their results. Chapters 5 and 6 of this thesis give the details of the first Norne benchmark study from preparation to the results respectively; the analysis of the results was done by the author of this thesis. Through this study it has been found that the presence of the Norne benchmark case is of great importance in enhancing the quality of research in closed loop reservoir management with emphasis the use of various data type examples of time-lapse seismic data.