Abstract:

In order to explore the effective prediction method of coal seam in-situ stress, the Support Vector Machine (SVM) regression method was used to calculate the minimum horizontal principal stress. Combined with the other two directions of stress calculation, the in-situ geological stress model was constructed, and the three-dimensional in-situ stress field was visualized. The grey correlation method was used to screen out the logging parameters that are best correlated with the minimum horizontal principal stress, including caliper logging (CAL), compensated neutron logging (CNL), natural gamma-ray logging (GR), density logging (DEN), and the mean deep and shallow lateral resistivity logging (R). With these five training factors, the prediction model of the minimum horizontal principal stress was established with the SVM regression method. And the H3 well group in Hancheng block in the Eastern Erdos Basin was used as an example to calculate the coal seam stress. The results indicate that the in-situ stress in three directions of the study area has an increasing trend with increasing burial depth, and the stress field also changes from the shallow geodynamic type to the deep geostatic type, and the stress environment of the coal reservoir correspondingly transformed from the extrusion to the extensional zone.

Key words: in-situ stress, support vector machine, logging, gray correlation, geological modeling