DATA ANALYTIC APPROACH IN PRODUCTION FORECAST AND OPTIMISATION FOR CONVENTIONAL OIL BROWNFIELD / THAM KEAT FU.
Material type:
TextPublication details: Kuala Lumpur : Asia Pacific University, 2021Description: 224 pages : illustrations ; 30 cmSubject(s): Reservoirs -- Computer simulation | Machine learningLOC classification: PM-32-35Dissertation note: A dissertation submitted to Asia Pacific University of Technology and Innovation in fulfilment of the requirements for the degree of Master of Science in Data & Business Analytics (APUMP1911DSBA(BI)(PR)). Summary: Production prediction or forecast is an important process in the upstream oil and gas industry, carrying financial significance and impact. Successful project execution and positive cash flow are dependent on recoverable reserves generated from accurate forecast. This is especially true for mature oilfields (brownfields) that are declining in production. The most used techniques in production forecast in oil brownfields are Numerical Reservoir Simulations (NRS) and Decline Curve Analysis (DCA). Despite their respective advantages, NRS is too time and effort consuming to have a properly history matched model that could be used for forecasting while DCA oversimplifies the physics behind reservoir fluid flow and the dynamic interactions of the producing conditions, resulting in deviating forecast. With the advancement of computing power and machine learning research, the abundant data gathered from the oilfields can be used to provide predictive analytics solutions in production prediction. In existing research, machine learning and deep learning models have been explored and compared with conventional methods such as NRS and DCA. The outcomes are promising and there is room for improvement. This study presents the methodology and results of predicting production rates of all producers from publicly available production and injection dataset obtained from Volve Field using regression models. The best performing model selected for the three production rates is coincidentally Extra Trees Regressor. The results obtained showed high correlation coefficient of approximately 99% in both validation and testing datasets for oil, gas, and water rates prediction. Apart from that, the other evaluation metrics such as Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE) obtained are comparable and some better than in literatures
| Item type | Current library | Collection | Call number | Copy number | Status | Date due | Barcode |
|---|---|---|---|---|---|---|---|
Reference
|
APU Library Reference Collection | Masters Theses | PM-32-35 (Browse shelf (Opens below)) | 1 | Not for loan (Restricted access) | 00019065 |
Browsing APU Library shelves, Shelving location: Reference Collection, Collection: Masters Theses Close shelf browser (Hides shelf browser)
A dissertation submitted to Asia Pacific University of Technology and Innovation in fulfilment of the requirements for the degree of Master of Science in Data & Business Analytics (APUMP1911DSBA(BI)(PR)).
Production prediction or forecast is an important process in the upstream oil and gas industry, carrying financial significance and impact. Successful project execution and positive cash flow are dependent on recoverable reserves generated from accurate forecast. This is especially true for mature oilfields (brownfields) that are declining in production. The most used techniques in production forecast in oil brownfields are Numerical Reservoir Simulations (NRS) and Decline Curve Analysis (DCA). Despite their respective advantages, NRS is too time and effort consuming to have a properly history matched model that could be used for forecasting while DCA oversimplifies the physics behind reservoir fluid flow and the dynamic interactions of the producing conditions, resulting in deviating forecast. With the advancement of computing power and machine learning research, the abundant data gathered from the oilfields can be used to provide predictive analytics solutions in production prediction. In existing research, machine learning and deep learning models have been explored and compared with conventional methods such as NRS and DCA. The outcomes are promising and there is room for improvement. This study presents the methodology and results of predicting production rates of all producers from publicly available production and injection dataset obtained from Volve Field using regression models. The best performing model selected for the three production rates is coincidentally Extra Trees Regressor. The results obtained showed high correlation coefficient of approximately 99% in both validation and testing datasets for oil, gas, and water rates prediction. Apart from that, the other evaluation metrics such as Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE) obtained are comparable and some better than in literatures
There are no comments on this title.