A new criterion for predicting liquid loading in vertical gas wells

• Liquid film reversal criterion is too conservative for predicting liquid loading based on experimental observation and analysis. • A new criterion for liquid-loading onset was established by combining the nodal analysis system and the new critical point criterion. • The new proposed method can better explain various field symptoms of gas wells than liquid-droplet and liquid-film reversal models. Predicting liquid-loading onset is a challenging subject as part of the analysis of gas wells. Although many experimental investigations show that the liquid film reversal leads to liquid loaded and this criterion has been used in a number of models, these models fail to ensure reliable prediction accuracy when predicting liquid-loading onset in vertical gas wells. Flow pattern is an indicator reflecting production characteristics in vertical gas well. In this study, we aim to review these methods to predict liquid loading and compare experimental measurements at liquid film reversal with liquid-loading field symptoms. For analyzing the static characteristics and dynamic change rules of different flow patterns, liquid holdup and transient pressure fluctuation are used as evaluation indexes. With the blocked wellbore as the critical point, the churn flow is innovatively split into pseudo-annular flow and pseudo-slug flow. Then, a new criterion for predicting liquid loading based on liquid-film movement at different gas velocities is proposed. We have combined the nodal analysis system and new critical-point criterion to establish a more accurate model to calculate liquid loading. The model of the new critical-point criterion is derived by applying the momentum equation of liquid film when liquid flow rate of downflow and upflow in the laminar layer are both equal to half of the liquid flowrate at the inlet. The proposed method can better explain various field symptoms of gas wells. Furthermore, by comparing with field data, the results show that the new model has improved predictive accuracy compared to existing models.