数据驱动的水平管气液两相流压降预测方法OA

The accurate prediction of pressure drop in horizontal gas-liquid two-phase flow has important guiding significance for productivity evaluation of horizontal wells,optimization design of surface pipelines,and flow safety assurance of subsea pipelines.In this study,through conducting laboratory simulation experiments of two-phase horizontal pipe flow under different pipe diameters and different gas-liquid flow rates,the traditional pressure drop calculation methods such as Lockhart&Martinelli,Beggs&Brill,and Dukler were evaluated.Meanwhile,machine learning methods such as support vector machine,random forest,and BP neural network were used to train and perform regression prediction on the experimental data,thereby establishing a data-driven pressure drop prediction model for horizontal gas-liquid two-phase flow.Results show that compared with the classical models,the data-driven model has higher prediction accuracy.Among the three types of machine learning methods,the horizontal pipe pressure drop model established based on the BP neural network algorithm has the highest prediction accuracy,which can meet the demand for horizontal pipe pressure drop prediction within the studied working conditions.This study has confirmed the potential of the data-driven method in predicting the pressure drop of gas-liquid two-phase flow and can be extended to other multiphase flow application scenarios.