Selecting and Modifying Multiphase Correlations for Gas-Lift Wells Using Machine Learning Algorithms

In the conditions of the development of the oil fields of the Cuu Long Basin in the continental shelf of the Republic of Vietnam, in the absence of downhole gauge systems, the urgent task is improving the accuracy of the calculation of bottomhole pressure in the producing wells based on the operation modes and construction. The aim of the paper is to create tools for selecting and modifying the correlation of multiphase flow most suitable for the development of a particular group of fields, as well as to develop a tool to implement effective management of the modes of operation of gas-lift wells by choosing the optimal gaslift injection rate. Based on data from 814 instrumental measurements in wells with different construction, liquid flow rate, watercut, GOR and gaslift injection, the calculation of bottom hole pressures was made. The calculated and actual bottomhole pressures were compared with five correlations of multiphase flow, the most suitable correlations were determined and modified, including using machine learning methods, which helped to significantly improve the convergence of calculated and actual bottomhole pressures. On the basis of the newly modified correlation, a calculation of bottom hole pressure (BHP) in each production well was made, the calculation of the change in bottomhole pressure when changing the operating modes of wells has been implemented. For the field group of the Cuu Long Basin, it was revealed that with the increase in watercut in the producing wells significantly reduces the efficiency of the gas-lift method of operation. This effect is not reflected in the widespread correlations of multiphase flow, which does not allow to use the results of calculations without making additional edits. A way to adapt the calculation values to instrumental measurements has been implemented, one of the known correlations has been modified and used in the forecast of changes in bottomhole pressure after changes in operating modes of wells throughout the well stock.