Abstract: Taking 12 casing-deformed horizontal wells in the Permian Lucaogou Formation of the Jimsar Sag as the research object，based on three-dimension seismic attribute data，a comparative analysis is conducted to examine the correspondence between casing deformation locations induced by geological factors and the development of natural fractures. A risk prediction model for casing deformation locations is established using the Random Forest algorithm from machine learning. The result shows that：（1）Casing deformation occurs primarily at locations where natural fracture zones develop and at boundaries of reservoir heterogeneity，with a nonlinear correlation between the two.（2）The number of decision trees and the number of features for node splitting in the Random Forest casing deformation risk prediction model are optimized using grid search and 5-fold crossvalidation. Considering both computational accuracy and efficiency，the optimal number of features for node splitting is set to 2，and the number of decision trees is set to 100.（3）The application of actual data shows that the Random Forest algorithm achieves a prediction accuracy of 87.85% for casing deformation locations in the tight oil reservoirs of the Permian Lucaogou Formation in the Jimsar Sag. The risk prediction results output by the model can provide guidance for optimizing fracturing designs and adjusting construction parameters.

Key words: machine learning, random forest algorithm, hydraulic fracturing, casing deformation, Lucaogou Formation, Permian, Jimsar Sag