Search in this Book
sort by
استفاده از روش‌های ادغام داده برای جریان‌های چندفازی در شبکه‌های متخلخل
299 viewed

استفاده از روش‌های ادغام داده برای جریان‌های چندفازی در شبکه‌های متخلخل

نجفی، حسین Najafi, Hossein

Use of Data Assimilation Methods for Multiphase Flow in Porous Media

Najafi, Hossein | 2020

299 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 53493 (46)
  4. University: Sharif University of Technology
  5. Department: Energy Engineering
  6. Advisor(s): Rajabi Ghahnavieh, Abbas; Bazargan, Hamid
  7. Abstract:
  8. The importance of optimizing the extraction process of available resources increases each day due to the increasing energy consumption and the lack of energy resources. Oil and gas are one of the most important sources of energy. Although existing oil and gas resources are thought to be sufficient to meet the growing energy demand for the next few decades, given the non-renewable nature of these resources and the growing demand for oil and gas, it will become much harder to meet the future energy demand. Many existing oil fields are now in the process of maturing, and the discovery of large new oil fields is rare. As a result, new technologies must be used in the future to meet this demand, assuming no new oil fields are discovered and the current rate of decline is stable. Optimal operation and management of oil and gas reservoirs has a great impact on the production rate and lifetime of reservoirs. One of the most important issues in reservoir management is the History Matching process. History Matching is the process of modifying an initial reservoir model to match production data so that the reservoir model matches the observation data. Traditional methods for doing such a process are time consuming and costly. As a result, today new methods called Monte Carlo are used in the Bayesian framework. In this thesis, the Population Markov Chain Monte Carlo(POP-MCMC) and Ensemble Kalman Filter(EnKF) methods will be used to perform the History Matching process. This method is highly efficient compared to the traditional Markov chain Monte Carlo method and significantly reduces computational cost. Generating 20,000 samples of the posterior density function in the history matching process using the POP-MCMC method requires 64% less time than the traditional MCMC method. The POP-MCMC method consists of three main parts: Mutation, Crossover and Exchange. Due to the characteristics of the Crossover part, the process of sampling variables is not trapped in local optima, thus increasing the space search efficiency of variables. The EnKF method, despite its high speed and much lower computational cost than the two methods mentioned, has good accuracy. By modifying the EnKF method, we increased the accuracy of the model predictions. In addition, in modeling a reservoir, it is divided into thousands of blocks, so to solve a History Matching problem, the high dimensions of the problem are one of the most important challenges. Therefore, in this study, we used Karhunen-Loève theorem to reduce the dimensions. In this research, PUNQ-S3 was used to implement the methods and compare them with each other
  9. Keywords:
  10. History Matching ; Monte Carlo Method ; Markov Chain ; Bayesian Framework ; Markov Chain Monte Carlo ; Ensemble Kalman Filter ; Sampling ; Reservoir Management

 Digital Object List