Аңдатпа
Background: The outcome of a geological and technical measure is governed by the combined influence of reservoir properties and operating conditions before treatment. Historical field data can support candidate-well selection, but the transfer of established relationships to another field remains uncertain when the sample structure, out-of-sample validation, and applicability limits are not stated explicitly. Aim: To analyse and evaluate the influence of geological, physical, and technological reservoir characteristics on five geological and technical measure performance indicators and to determine whether the resulting models can support a transparent comparative choice among treatment options. Materials and methods: The dataset comprised 107 historical intervention cases: 38 near-wellbore-zone treatments with perforation, 32 acid treatments with perforation, and 37 production-optimization cases. Sixteen input indicators were transformed into nine physically interpretable composite variables. Separate multiple linear regressions were fitted for each intervention type and each of five outcomes. Internal out-of-sample performance was assessed using shuffled five-fold cross-validation with a fixed random seed of 42. Transferability was examined for five West Prorva reservoir units with complete input data by calculating three alternative scenarios per unit and applying fuzzy max-min ranking. Results: For near-wellbore-zone treatment with perforation, cross-validated R² was 0.530 for effect duration and 0.522 for post-treatment oil rate. For acid treatment with perforation, the highest value was 0.667 for post-treatment oil rate. Production-optimization models were unstable except for a weak positive result for post-treatment oil rate (R² = 0.235). In the West Prorva comparison, production optimization ranked first for units IX-2 and VIII-2, near-wellbore-zone treatment with perforation for T-III and VIII-3, and acid treatment with perforation for T-II. Decision-membership function values were low to moderate (0.15–0.48). Conclusion: The approach is suitable for preliminary screening and comparative decision support, but not for unrestricted prediction outside the calibration domain. Local recalibration and validation against observed post-intervention results are required before operational use.