Abstract: An accurate estimation of the flow pattern of a horizontal gas well that produces water is the key to predicating the pressure drop of the wellbore and establishing a sensible water-drainage production plan. The angle of a horizontal well constantly varies from 90 ° to 0 according to the flow direction. A unified flow pattern map that describes the two-phase flow of horizontal wells has not yet been discovered. Therefore, three separate flow pattern maps are used to describe the horizontal, slanted, and vertical pipes individually. These maps are obtained under extremely different experimental conditions. In addition, the gas wells may produce extremely low water output. In such cases, the gas-liquid ratio may exceed the valid coordinate range of the commonly used gas-liquid two-phase flow pattern map leading to significant errors in the prediction results. Considering all these drawbacks, this study has developed an air-water two-phase flow simulation experimental device for horizontal, slanted, and vertical pipes. Provided the extremely high gas-liquid ratio of the gas well with water production, this study has conducted a gas-water two-phase flow pattern experiment using 7 slanted pipes and 641 horizontal pipes. It has summarized five flow patterns and the typical characteristics of horizontal gas wells. This work has used the dimensionless gas velocity number defined by Duns and Ros and the pipe inclination angle as the X-axis. A three-dimensional flow pattern map has been constructed to describe the gas-liquid two-phase pipe flow in the horizontal gas well. This study has also proposed a method to predict the flow pattern based on the back propagation neural network model. The pressure measurement results from 20 horizontal gas wells in the gas fields at Western Sichuan indicate that the flow map demonstrates an accuracy of 90% in prediction.

Key words: horizontal gas well, gas-liquid two-phase flow, pipe angle, three-dimensional flow pattern map, BP neural network