講演抄録/キーワード |
講演名 |
2018-11-05 14:20
A Study on a Feature Based Clustering and Decision Tree Regressions for Estimating the Bubble Point Pressure of Crude Oils ○Meshal Almashan・Yoshiaki Narusue・Hiroyuki Morikawa(The Univ. of Tokyo) ASN2018-68 |
抄録 |
(和) |
Bubble point pressure (Pb) is one of the most important Pressure-Volume-Temperature (PVT) properties of any crude oil system, it is required in calculations used in production and reservoir engineering. By laboratory experiments, the PVT properties can accurately be determined. However, laboratory experiments require applying special tests on the oil samples which need to be handled with care. As alternative approaches, researchers have developed equations of state (EOS) and empirical correlations for estimating the PVT properties. However, these alternative solutions have some limitations. With the introduction of the machine learning applications in the petroleum industry, other researchers have studied the predictive power of several machine learning models for estimating Pb. One of the most commonly tested and applied modeling schemes are the artificial neural networks (ANNs). However, ANNs suffer from the ?black-box? problem and there is no direct and heuristic way of determining the importance of the input parameters in predicting the PVT properties. In the present study, a Boosted Decision Tree Regression (BDTR) predictive model with K-means clustering is built and evaluated in the estimation of Pb. |
(英) |
Bubble point pressure (Pb) is one of the most important Pressure-Volume-Temperature (PVT) properties of any crude oil system, it is required in calculations used in production and reservoir engineering. By laboratory experiments, the PVT properties can accurately be determined. However, laboratory experiments require applying special tests on the oil samples which need to be handled with care. As alternative approaches, researchers have developed equations of state (EOS) and empirical correlations for estimating the PVT properties. However, these alternative solutions have some limitations. With the introduction of the machine learning applications in the petroleum industry, other researchers have studied the predictive power of several machine learning models for estimating Pb. One of the most commonly tested and applied modeling schemes are the artificial neural networks (ANNs). However, ANNs suffer from the ?black-box? problem and there is no direct and heuristic way of determining the importance of the input parameters in predicting the PVT properties. In the present study, a Boosted Decision Tree Regression (BDTR) predictive model with K-means clustering is built and evaluated in the estimation of Pb. |
キーワード |
(和) |
Predictive model / PVT / Oil and gas / Reservoir characterization / / / / |
(英) |
Predictive model / PVT / Oil and gas / Reservoir characterization / / / / |
文献情報 |
信学技報, vol. 118, no. 282, ASN2018-68, pp. 75-80, 2018年11月. |
資料番号 |
ASN2018-68 |
発行日 |
2018-10-29 (ASN) |
ISSN |
Online edition: ISSN 2432-6380 |
著作権に ついて |
技術研究報告に掲載された論文の著作権は電子情報通信学会に帰属します.(許諾番号:10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
PDFダウンロード |
ASN2018-68 |
|