Identification of oil-water flow patterns using conductance probe in vertical well

Abstract

In this paper, a sensor of conductance probe is proposed to detect the electrical characteristics of the oil-water flow in vertical well. Statistic and wavelet packet decomposition are employed to extract the features of the voltage response of conductance probe. A method based on principal component analysis (PCA) and support vector classification (SVC) is proposed to identify the flow patterns from the water-in-oil, transition, and oil-in-water flow patterns. Experiments were carried out in a 125 mm vertical well within the flow rate range of 10~200 m3/d and the water content range of 10~90% in Daqing Oilfield, China. Experimental results reveal that the optimal identification accuracy of training set is obtained as 100%, and that of testing set is achieved as 96.25%. Corresponding quantity of of principal component is 7, and cross validation accuracy is 95%. Consequently, the proposed method is feasible and effective to identify the flow patterns of oil-water flow using conductance probe sensor in vertical well.