A Comparison between Different Uncertainty Quantification Methods ImplementedOver a Benchmark Reservoir Model Aiding Assisted History Match Process

سال انتشار: 1395
نوع سند: مقاله کنفرانسی
زبان: انگلیسی
مشاهده: 742

فایل این مقاله در 11 صفحه با فرمت PDF قابل دریافت می باشد

استخراج به نرم افزارهای پژوهشی:

لینک ثابت به این مقاله:

شناسه ملی سند علمی:

CHECONF03_403

تاریخ نمایه سازی: 14 آذر 1395

چکیده مقاله:

Because of the measurement and modeling errors incorporated into the simulation modeling, estimated petrophysical properties are always associated with uncertainty. Quantifying this uncertainty plays an important role at the political and economic decisiontaken by the company managers. Hence, three different Uncertainty Quantification (UQ) techniques are employed at this study for uncertainty quantification of the Teal South benchmark reservoir and their performance are compared accordingly. These three differentmethods include: LMAP (Linearization about the MAP), RML (Randomized Maximum Likelihood), and McMc (Markov Chain Monte Carlo). Since these UQ techniques are based on the optimization of the reservoir parameters for certain well history data, Gauss-Newton(GN) with line search technique and Restricted Step-Size method were used as the optimization algorithms within the UQ techniques. Totally, three petrophyscial parameters, Kx, Ky, and porosity, at each grid block were used for UQ purpose. Applying the differentUQ methods over the benchmark model, it was observed that the CDF (Cumulative probability Density Function) plot of the ultimate recovery obtained by the RML technique is a combination of the CDF plots achieved by using the McMc and LMAP techniques.Consequently, the PDF plot of the RML revealed a bimodal normal distribution contrary to the LMAP and McMc techniques. Using the Maximum Likelihood Estimation (MLE) over the PDF plot of the RML method to fit a bimodal normal distribution, it was observed that thefirst estimated mean value is more close to the mean of the PDF of the LMAP method while the second mean value of the bimodal PDF is much more close to the mean obtained by using the McMc method. Also, while GN with line search algorithm is sufficient for LMAP andMcMc techniques, this algorithm is not converged for the RML method and Restricted Step- Size method was used within the RML technique. Analyzing the P10, P50, and P90 of the ultimate recovery data, it was also concluded that the McMc leads to the highest P10, P50,and P90 values. Furthermore, RML ends up with the least P10 and P50 values while the LMAP technique leads to the lowest P90

نویسندگان

Meisam Adibifard

Petroleum Engineering Department, Amirkabir University of Technology

Mohammad Ahmadi

Petroleum Engineering Department, Amirkabir University of Technology

Alireza Kazemi

Petroleum Engineering Department, Amirkabir University of Technology

مراجع و منابع این مقاله:

لیست زیر مراجع و منابع استفاده شده در این مقاله را نمایش می دهد. این مراجع به صورت کاملا ماشینی و بر اساس هوش مصنوعی استخراج شده اند و لذا ممکن است دارای اشکالاتی باشند که به مرور زمان دقت استخراج این محتوا افزایش می یابد. مراجعی که مقالات مربوط به آنها در سیویلیکا نمایه شده و پیدا شده اند، به خود مقاله لینک شده اند :
  • Katterbauer, K., Arango, S., Sun, S. and Hoteit, I., 2015. ...
  • Maschio, C. and Schiozer, D.J., 2005. Assisted history matching using ...
  • Maschio, C. and Schiozer, D.J., 2014. Bayesian history matching using ...
  • Mohamed, L, Christie, M.A. and Demyanov, V., 2010. Comparison of ...
  • Oliver, D.S., Reynolds, A.C. and Liu, N., 2008. Inverse theory ...
  • Santhosh, E.C. and Sangwai, J.S., 2016. A hybrid differential evolution ...
  • Sarma, P., Durlofsky, L. and Aziz, K., 2008. Computational techniques ...
  • Sultan, A., Ouenes, A. and Weiss, W., 1994. Automatic history ...
  • نمایش کامل مراجع