The geomechanical environment in the complex structural area of southern Sichuan is complex and the casing deformation occurs frequently, which leads to the difficulty of horizontal well fracturing transformation, resulting in the failure of downhole engineering measures and the huge loss of single well production, thus greatly affecting the production and prolonging the production reduction cycle. However, the geomechanical factors in complex structural areas are closely related to the risk of casing deformation. Therefore, it is necessary to clarify the relationship between the geomechanical parameters in local structures and the risk of casing deformation, and to construct the corresponding evaluation model to accurately predict the risk of casing deformation. The basic characteristics of casing deformation in deep shale gas wells were studied by using the existing microseismic monitoring, casing geometry monitoring and multi-arm caliper logging data, and the factor series of quantitative evaluation of casing deformation risk was constructed based on the amount and degree of casing deformation. Combined with the existing multi-level fracture system interpretation data, in-situ stress field simulation results and drilling data. The relationship between casing deformation risk factor series and reservoir geomechanical parameters was studied, and the intelligent classification and identification algorithm of casing deformation risk in deep shale gas wells was established, which is applied to the early warning of casing deformation risk before new drilling pressure. The results show that the risk level of casing deformation in deep shale gas wells in southern Sichuan can be divided into five grades based on the degree and characteristics of casing deformation. The greater the grade, the greater the degree of casing deformation and the greater the damage. The degree of casing deformation is mainly closely related to the geomechanical parameters of deep reservoirs. In areas with strong tectonic deformation (curvature > 0.25 km^-1), large fault dip angle (> 45°), small angle between fault strike and maximum horizontal principal stress direction (18 MPa), and local compressive strike-slip stress state (structure index > 1.65), casing deformation of grade 3 or above is prone to occur, and the damage intensity of casing deformation is large. Based on the relational database between the casing deformation factor series and the reservoir geomechanical parameters, the Bayesian classification and identification algorithm is used to establish the casing deformation risk prediction analysis model in the study area and successfully applied to the new horizontal well. The prediction accuracy is above 90%, which can be popularized and applied. This method can reduce the loss of horizontal well section before fracturing, improve the success rate of horizontal well fracturing, and improve the development effect and economic benefits of deep shale gas wells.