Optimized Dynamic Classification of Offshore Gas Wells Using Principal-Factor Screening and Cluster Analysis

A gas field in the eastern South China Sea has entered the middle-to-late stage of development and is strongly affected by medium-to-strong water drive and severe edge- and bottom-water encroachment. After water breakthrough, gas wells exhibit different productivity-decline patterns, and some wells suffer from insufficient liquid-carrying capacity, liquid loading, and production decline. Conventional classification methods based on individual indicators or empirical thresholds cannot adequately characterize the current production state, stable-production behavior, and future deterioration risk. Therefore, an ELF-index-based HGC-ELF Pro framework was developed for dynamic gas-well classification and deliquification-process screening. PCA and CRITIC retained 14 indicators from 22 candidates, and the first five principal components explained 90.089% of the total variance. K-means divided 25 wells into three classes; K = 3 produced the highest reported average silhouette coefficient of 0.617 among K = 2–6. Class I contained 12 wells (48%), Class II contained 6 wells (24%), and Class III contained 7 wells (28%). The recommended primary processes included foam-assisted deliquification for 13 wells (52%), gas lift/ESP screening for 6 wells (24%), gas lift for 4 wells (16%), and velocity strings for 2 wells (8%). Because independent field class labels are unavailable, supervised classification accuracy is not reported. The framework integrates multi-timescale production information, interpretable E/L/F state representation, cluster-validity assessment, and differentiated process screening. The results provide a quantitative basis for hierarchical well management, although the small sample and field-specific calibration requirements limit direct generalization.