The borehole wall stability of depleted reservoirs is a key problem that restricts the deeper potential exploitation of old oilfields. Based on these, according to the seepage and elastic mechanics theory of porous media, combined with laboratory rock mechanics experiments, the dynamic evaluation model of geomechanical parameters of the old reservoir is established, and on this basis, the evaluation method of wellbore stability of new wells in the old reservoir is established, and the changing laws and influencing factors of the wellbore stability of the old reservoir are quantitatively evaluated. The research shows that reservoir pressure, reservoir porosity, effective stress coefficient, strength, in situ magnitude, and direction were all changed due to long-term exploitation; both collapse and fracture pressure decreased, the loss risk of drilling fluid was increased, and safe drilling azimuth was changed. The results can guide reasonable choices such as well position, track and mud density, and strengthening sealing while drilling in order to ensure wellbore stability effectively.

Cross-well lithology recognition is important for reservoir characterization, but its application to newly drilled wells is constrained by inter-well variations in logging responses and limited lithological labels. This study proposes a geoscience knowledge-guided machine-learning method for lithology recognition under limited target-well labels. Geochemical indicators and conventional logging data are combined to construct geologically interpretable features, while few-shot transfer learning is used to reduce class imbalance and inter-well distribution differences. Geological knowledge is introduced as probabilistic constraints and integrated with model predictions through uncertainty-aware fusion. The method is evaluated using volcanic reservoir well-logging data from the Pearl River Mouth Basin. Three representative wells are used as the source domain and an independent newly drilled well as the target domain, with three labeled samples per lithology class used for adaptation and the remaining samples for independent testing. Repeated experiments with 10 random seeds yield an accuracy of 88.03% ± 6.09%, with improved classification performance and lower variability than the compared baseline methods. The results show that combining geological knowledge with few-shot learning can improve the robustness of cross-well lithology recognition and provide a practical approach for lithology prediction in heterogeneous volcanic reservoirs.