Guangqiang Cao

Plunger lift systems are vital for optimizing production in gas wells, but their performance can be compromised by various operational anomalies. Traditional diagnostic methods and conventional convolutional neural network (CNN) approaches often struggle with the complex, transient data from these systems, particularly in capturing long-range temporal dependencies and generalizing from limited, imbalanced datasets. This study presents an enhanced diagnostic framework for plunger lift anomaly detection by leveraging the strengths of a pre-trained Vision Transformer (ViT). The methodology transforms one-dimensional time-series pressure data into two-dimensional image representations using the element-wise summation of Gramian Angular Summation Field (GASF) and Gramian Angular Difference Field (GADF), which simultaneously preserves global operational trends and local transient dynamics for vision model analysis. The ViT model, initialized with pre-trained weights, is further optimized using Bayesian optimization (BO) for hyperparameter tuning, and a tailored data augmentation pipeline is employed to improve robustness. Comparative evaluations demonstrate that the proposed ViT-based approach, particularly the ViT + GAF + BO model, significantly outperforms baseline CNN models and their optimized variants, achieving the highest Precision, Recall, and F1-score, with an F1-score of 0.93. Visualizations using t-SNE confirm the ViT’s superior capability in learning discriminative features, showcasing well-separated clusters for different operational conditions compared to CNNs. This research underscores the potential of pre-trained ViTs combined with appropriate data representation and optimization techniques for achieving accurate and reliable anomaly detection in plunger lift systems.

Pressure gradient prediction is crucial in gas well analysis. The experiment is the most effective method of understanding the flow characteristics in horizontal gas wells. The greatest difference between experimental and high-pressure conditions is gas density, which could cause the established multiphase correlations unreliable when they are applied to high-pressure gas wells. Similarity numbers are widely employed in predicting flow behavior. Nevertheless, few studies focused on this area. In addition, gas wells are characterized as high gas–liquid ratio; the majority empirical correlations were developed for oil wells, which have more consideration in low gas–liquid ratio, influencing the precision of gas well models. An experimental examination of gas–liquid flow has been carried out in this study. First, the experimental test matrix was designed to meet each flow pattern. Next, the effect of gas velocity, liquid velocity, pipe diameter, water-cut, and inclined angle on liquid holdup was explored. Subsequently, the similarity numbers suggested have been investigated and assessed for pressure scaling up. Finally, a comprehensive model was established, which was developed to forecast pressure gradient in gas wells. Field data were supplied to assess the new correlation. The results demonstrated that the Duns–Ros and the modified Duns–Ros dimensionless numbers were improper for pressure scaling up, whereas the Hewitt–Robert Number performs best. Based on the field data, the new correlation with Hewitt–Robert Number was superior to extensively employed pressure drop correlations, showing that it can deal with predicting pressure gradient in gas wells.

In order to explore the feasibility of the Computational Fluid Dynamics (CFD) method for plunger gas lift and the flow pattern in a long wellbore, a lab simulation test was conducted to verify the reliability of the CFD simulation results. Verification data were provided for CFD simulation through carrying out a plunger gas lift lab test for vertical wells, identifying the plunger movement patterns and delivery rates under different flow pressures, and determining the minimum plunger startup pressure or differential pressure and leakage, and an identical CFD physical model was created on the basis of the lab test model, CFD dynamic grid programming and CFD simulation were conducted under test conditions, and a comparison of the simulation and test results was made to identify the calculation accuracy and the rationality of the CFD model and method; finally, boundary conditions such as temperature and pressure were set according to the actual long wellbore (200 m) conditions, the CFD simulation was performed, and the impacts of the downhole conditions on the gas lift performance were analyzed, so as to develop CFD calculation methods to predict the bottom-hole flow pressure, plunger speed, and delivery rate. The results show: the average plunger speed range is 7.74–22.5 m/s when the flow pressure varies from 199.77 to 632.93 kPa, and the leakage rate increases in a nearly linear way with the speed; compared with the lab test results, the simulation results from the created dynamic grid model and multi-phase turbulent flow model have the leakage error of 7.2% and the plunger speed average error is smaller than 11.1%; under long-wellbore conditions, the plunger lift speed shows the change pattern of increasing and then decreasing, the wellbore pressure has a wave-like drop, and in addition to this pressure drop characteristic, the fact that the plunger startup pressure differential increases with the wellhead pressure should be considered (when the bottom-hole pressure is 15 MPa, the wellhead pressure must not exceed 10 MPa).