Prediction of Two-Phase Flow Regime in Oil Wells Using Hybrid Models

Abstract

This study investigates the application of hybrid machine learning techniques, including boosting, bagging, voting, and stacking for flow regime prediction in twophase vertical pipe flow. We propose a decision tree-based ensemble classifier utilizing algorithms like Random Trees (RT), J48, Reduced Error Pruning Trees (REPT), Logistic Model Trees (LMT), and Decision Trees with Naive Bayes (NBT).

The effectiveness of the chosen hybrid algorithm was assessed using a comprehensive suite of metrics, including classification accuracy, precision, recall, confusion matrix, F1-score, and PRC area. Our investigation revealed that ensemble methods, particularly boosting (AdaBoost, LogitBoost, MultiBoosting) and, achieved superior prediction accuracy compared to individual classifiers. Notably, MultiBoosting exhibited the most promising performance within the boosting category. These findings conclusively demonstrate the superiority of ensemble algorithms over single classifiers in predicting flow regimes. By leveraging this approach, the accuracy of flow regime prediction was demonstrably increased, reaching a level as high as 96%.

The study introduces a superior method for predicting two-phase flow regimes in vertical flows, achieving high accuracy and reducing complexity and cost, resulting in reliable results under various operating conditions.