Comparison of machine learning methods for multiphase flowrate prediction

Abstract

In this paper, three prevailing machine learning methods, i.e. Deep Neural Network (DNN), Support Vector Machine (SVM) and Gradient Boosting Decision Tree (GBDT) models were investigated and compared to estimate the flowrate of oil/gas/water three-phase flow. The time-series differential pressure signals collected from Venturi tube together with pressure and temperature measurements were utilized as input. Multiphase flow experiments were conducted on a laboratory-scale multiphase flow facility. Experimental results suggest that DNN and SVM based methods were able to achieve accurate and reliable estimation of multiphase flowrate, whilst GBDT failed to fit the estimation process well. Another finding emerged from this study is that volumetric gas phase flowrate can also be accurately predicted by implementing SVM model.