SURROGATE RESERVOIR MODEL EXTRACTION FOR MULTIPHASE FLOW SIMULATION
Reservoir simulation softwares are used as an important tool in oil and gas industries to predict the responses of the reservoir. Due the large number of grid blocks and hetero geneity in the reservoir model, large number of simulations are required to narrow down the risk in reservoir productivi...
| Main Author: | MEMON, PARAS QADIR |
|---|---|
| Format: | Thesis |
| Language: | English |
| Institution: | Universiti Teknologi Petronas |
| Record Id / ISBN-0: | utp-utpedia.21468 / |
| Published: |
2015
|
| Subjects: | |
| Online Access: |
http://utpedia.utp.edu.my/21468/1/2015-INFORMATION%20TECHNOLOGY-SURROGATE%20RESERVOIR%20MODEL%20EXTRACTION%20FOR%20MULTIPHASE%20FLOW%20SIMULATION-PARAS%20QADIR%20MEMON-MASTER%20OF%20INFORMATION%20TECHNOLOGY.pdf http://utpedia.utp.edu.my/21468/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: |
Reservoir simulation softwares are used as an important tool in oil and gas industries to
predict the responses of the reservoir. Due the large number of grid blocks and hetero
geneity in the reservoir model, large number of simulations are required to narrow down
the risk in reservoir productivity. To mitigate this problem, Surrogate Reservoir Model
(SRM) is considered as a potential solution to reduce simulation time. The SRM is
used to predict the average reservoir pressure, production rate and bottom-hole flowing
pressure (BHFP) based on the time complexity. The main objective of this research is
to develop dynamic well Surrogate Reservoir Model (SRM), that mines the output data
from a conventional reservoir simulator. Key input parameters e,g. porosity, perme
ability are identified from reservoir model using principal component analysis (PCA)
technique. Two supervised Artificial Neural Network (ANN), i.e. backpropagation neu
ral network (BPNN) and radial basis neural network (RBNN) is used to build SRM for
system prediction. Mean Square Error (MSE) is used to calculate the error between the
target and predicted output in order to select the SRM with minimum error value. The
RBNN is shown to be the more effective in the development of SRM. |
|---|