Identification of Oil-Water Flow Patterns in a Vertical Well Using a Dual-Ring Conductance Probe Array

In this paper, a dual-ring conductance probe array-based well logging instrument was developed and a method based on the support vector classification (SVC) and voting methods was proposed to identify the oil-water flow patterns in a vertical well. The patterns of the oil-water flow in a vertical well are classified into five patterns, i.e., pure oil phase, pure water phase, water-in-oil, oil-in-water, and transition. The conductance probe records the time-varying electrical characteristics of the oil-water flow, which is referred to as the original signals. Various features are extracted to characterize each original signal. The features are first treated through principal component analysis (PCA) to decrease data redundancy in the original features. A nonlinear SVC model is then established to map the PCA-treated features into a flow pattern. To identify the flow pattern, it can be voted by an individual probe or by probe combinations. Experiments were carried out in a vertical pipe with an inner diameter of 125 mm and a height of 24 m on the industrial-scale experimental multiphase flow setup in Daqing Oilfield, China. In the experiment, the oil-water two-phase flow was tested and the total flow rate was varied from 10 to 200 $\text{m}^{3}$ per day, equivalent to 0.0094 to 0.1886 m $\cdot ~\textrm {s}^{-1}$, and the water cut was varied from 0% to 100%. The results obtained demonstrate that the developed probe array-based instrument can increase the reliability of flow pattern recognition, compared with the single probe-based instrument. The identification accuracy obtained using the optimal probe combination and the proposed method is 97.95% ± 2.01% (mean ± std), and higher than that obtained using a single probe and the SVC model.