Mechanism‐Data Coupling Driven Prediction Models of Liquid Leakage for Plunger Lift in Inclined Tube

ABSTRACTPlunger lift is a commonly used drainage gas production technology that can efficiently discharge bottomhole loaded liquid, but the inevitable liquid leakage between the plunger and the tubing wall has a significant negative impact on the drainage efficiency. Accurate prediction of liquid leakage rate is crucial for the design and optimization of plunger lift processes. The existing liquid leakage models are built for vertical wells and cannot meet the application requirements of inclined directional wells. In this study, plunger lift experiments were conducted at five different inclination angles to investigate the effects of plunger velocity and starting pressure difference on the liquid leakage. Furthermore, three conventional liquid leakage mechanism models were evaluated and revised, and a data model for predicting liquid leakage rate based on BP neural network and two mechanism‐data coupling models under input variable constraints and ensemble optimization methods were established. The results indicate that there is a nearly linear relationship between liquid leakage rate and plunger velocity in both vertical and inclined pipes, and the conventional mechanism models predict significantly higher liquid leakage rate than the measured values. After introducing dimensionless correction factor, the prediction accuracy of the modified mechanism model was significantly improved, with an R2 of 0.939. In addition, the data‐driven BP neural network model has also achieved good prediction accuracy. Moreover, incorporating mechanistic knowledge into the data model, such as input variable constraints and ensemble optimization, further enhances the prediction accuracy of liquid leakage rate, with R2 value of 0.99 for the testing data set. Therefore, developing mechanism‐data coupling models is an effective way to achieve reliable prediction of liquid leakage rate for plunger lift.