Water cut measurement of oil–water flow in vertical well by combining total flow rate and the response of a conductance probe. (12th August 2015)

• Record Type: Journal Article
• Authors: Chen, Jianjun; Xu, Lijun; Cao, Zhang; Zhang, Wen; Liu, Xingbin; Hu, Jinhai
• Is Part Of: Measurement science & technology. Volume 26:Number 9(2015:Sep.)
• Publication Date: 2015-08-12
• Subjects: multi-phase flow -- conductance probe -- water cut -- total flow rate -- support vector regression -- least squares regression; Physical measurements -- Periodicals; Scientific apparatus and instruments -- Periodicals; Equipment and Supplies -- Periodicals; Science -- instrumentation -- Periodicals; Technology -- instrumentation -- Periodicals; Mesures physiques -- Périodiques; Physical measurements; Scientific apparatus and instruments; Periodicals; 502.87
• DOI: 10.1088/0957-0233/26/9/095306
• Languages: English
• ISSNs: 0957-0233

Abstract

Abstract: In this paper, a conductance probe-based well logging instrument was developed and the total flow rate is combined with the response of the conductance probe to estimate the water cut of the oil–water flow in a vertical well. The conductance probe records the time-varying electrical characteristics of the oil–water flow. Linear least squares regression (LSR) and nonlinear support vector regression (SVR) were used to establish models to map the total flow rate and features extracted from the probe response onto the water cut, respectively. Principal component analysis (PCA) and partial least squares analysis (PLSA) techniques were employed to reduce data redundancy within the extracted features. An experiment was carried out in a vertical pipe with an inner diameter of 125 mm and a height of 24 m in an experimental multi-phase flow setup, Daqing Oilfield, China. In the experiment, oil–water flow was used and the total flow rate varied from 10 to 200 m 3 per day and the water cut varied from 0% to 100%. As a direct comparison, the cases were also studied when the total flow rate was not used as an independent input to the models. The results obtained demonstrate that: (1) the addition of the total flow rate as an input to the regression models can greatly improve the accuracy of water cut prediction, (2) the nonlinear SVR model performs much better than the linear LSR model, and (3) for the SVR model with the total flow rate as an input, the adoption of PCA or PLSA not only decreases the dimensions of inputs, but also increases prediction accuracy. The SVR model with five PCA-treated features plus the total flow rate achieves the best performance in water cut prediction, with a coefficient of determination ( R 2 ) as high as 0.9970. The corresponding root mean squared error (RMSE) and mean quoted error (MQE) are 0.0312% and 1.99%, respectively.