An interpretable XGBoost-based transfer learning framework for stress-sensitive pore volume compressibility prediction in carbonate rocks
Wensheng Luo
Pore volume compressibility (PVC) is a key petrophysical parameter controlling the mechanical response and production performance of reservoir rocks. However, the complex and heterogeneous pore systems of carbonate reservoirs make it difficult for traditional empirical models developed for sandstone to accurately predict PVC. In this study, an interpretable PVC prediction framework based on XGBoost and Bayesian optimization was proposed, and its generalization capability is further enhanced thro