با سلام خدمت کاربران در صورتی که با خطای سیستم پرداخت بانکی مواجه شدید از طریق کارت به کارت (6037997535328901 بانک ملی ناصر خنجری ) مقاله خود را دریافت کنید (تا مشکل رفع گردد).
ردیف | عنوان | نوع |
---|---|---|
1 |
Toward more realistic models of reservoir by cutting-edge characterization of permeability with MPS methods and deep-learning-based selection
به سمت مدل های واقع بینانه تر از مخزن با توصیف برش لبه نفوذپذیری با روش MPS و انتخاب مبتنی بر یادگیری عمیق-2019 Different sources of data are used to construct a reliable model of reservoir for oil/gas production. This model
ought to be matched with the production history of reservoir and also show reliable predictions for future
performance. To this end, permeability modeling (characterization of heterogeneity) is crucially important
which is proved to be done by Multiple Point Statistics (MPS) recently. Furthermore, deep learning methods are
massively used as a promising tool for regression applications. In this study, one MPS method is employed for
generating the reservoir realizations. Realizations, alongside their simulation outputs, are utilized for training a
convolutional deep network. In this manner, MPS is joined with deep learning to find the most appropriate
realization(s) of the reservoir based on the fluid flow simulation. Moreover, unseen MPS realizations as well as
another MPS realizations are used to verify the selection ability of trained network. The detailed architecture of
convolutional network is illustrated in this study.
The purpose of training this network and combination with MPS is to generate the matched realization(s) in
history period that also show acceptable reservoir behavior in the future times of reservoir simulation. After
training, the actual production data of selected realizations are obtained by simulation the reservoir for history
and also future times. The results show that selected realizations efficiently capture the trend of reference behavior.
Although these realizations lack identical permeability values, they have same texture of permeability
(permeability heterogeneity). Meanwhile, they show acceptable match in reservoir simulation outputs. By
proposed workflow, the uncertainty of permeability modeling is considered more exhaustively. It is done by
selecting the realizations from enormous possible realizations dataset and providing a deep learning tool which
is capable for screening quite large number of realizations. Interesting finding is satisfactory behavior of realization(
s) in both history and future periods of reservoir performance. Keywords: Reservoir simulation | Permeability modeling | Multiple point statistics | Deep learning | Geostatistic | Convolutional neural network (CNN) |
مقاله انگلیسی |