| Author | Do Van Thanh |
| Call Number | AIT RSPR no.PME-GEPG-12-03 |
| Subject(s) | Oil well logging--Cuu Long Basin (Vietnam)
|
| Note | A research study submitted in partial fulfillment of the requirements for the degree of Master Engineering ( Professional ) in Geo - Exploration and Petroleum Geo - Engineering |
| Publisher | Asian Institute of Technology |
| Series Statement | |
| Abstract | Kinh Ngu Trang (KNT) prospect has been discovered by the first exploration well (X - 01) with a reserve estimation of 43 MMbbls OIIP (5 - 7 MMbbls recoverable) from Early Oligocene E sequence, Tra Tan lower formation. It comprises two reservoir sections, E - upper and E - lower. The hydrocarbon reserve calculation is based on volumetric method using average porosity at the borehole , which might not be representative for whole prospect. U sing ANN analysis of both well log data and seismic attributes to predict porosity distribution for the whole reservoir is primary objective of this study. A seismic line passing through the X - 01 well location has been extracted from 3D seismic cube. It is a post - stacked seismic volume which was processed by pre - stack time migration technique. Grid size of this data is 12.5 m x 12.5 m in space direction and 2.0 milliseconds sample rate in vertical two - way time (TWT) direction/domain. T his domain has been chosen as reference scale when comparing well logs and the seismic data. Thus porosity curve needs to be converted to TWT space by using time - depth table from check shot survey. The reason is that predicted porosity distribution can the n be easily converted to depth domain whenever velocity functions are revised or updated. One additional step needs to do is resample porosity logs to the same sample rate as of seismic cube‟s for sample by sample comparison. It is because of seismic data existing at every 2 ms while the porosity log is available at very dense depth step and having also haphazard time interval after the conversion. Eight seismic attributes including reflection amplitude , instantaneous amplitude and its 1 st and 2 nd derivative , as well as instantaneous phase , frequency , dominant frequency and reflectivity were calculated and used as inputs for an ANN analysis while converted and re - sampled porosity log are used as desired output. The networks are first trained by different combination of the inputs in order to find how important to the network performance of each attribute would be. Secondly, it i s optimized by training with different number of the inputs as well as number hidden layers of the network. Some training functions such as Gradient Descent , Levenberg - Marquardt and Bayesian Regulation methods were tested. The other parameter s such as training rate, “logsig” , “tansig” and “purelin” transfer functions as well as data division methods were also implemented . But the network performance of these tests is not significant improved. The final porosity output using Bayesian Regulation training algorithm is the best match with highest R - value of 0.86. However, when applying to the whole seismic section it seems be over - fit training with abnormal porosities higher than 60% somewhere. This method should be further studied by analyzing more other input attributes combination as well as using all available boreholes in the studied area. |
| Year | 2013 |
| Corresponding Series Added Entry | |
| Type | Research Study Project Report (RSPR) |
| School | School of Engineering and Technology (SET) |
| Department | Department of Civil and Infrastucture Engineering (DCIE) |
| Academic Program/FoS | Geotechnical Engineering and Management (Master Engineering (Professional)) (PME - GEPG) |
| Chairperson(s) | Pham Huy Giao |
| Examination Committee(s) | Noppadol Phien - wej;Jang, Seonghyung |
| Scholarship Donor(s) | Fairfield Vietnam Co. , LTD HCM city, Viet Nam |
| Degree | Research studies project report (M. Eng.) - Asian Institute of Technology, 2013 |