Abstract: To improve the early identification of generator bearing temperature anomalies during wind turbine operation, the correlation between operating parameters and generator bearing temperature is investigated, and its application value for fault early warning is evaluated. Based on actual operation data collected by the Supervisory Control and Data Acquisition (SCADA) system from wind turbine generators, Spearman's Rank correlation coefficient is used to analyze the correlations between the main operation parameters, such as wind speed and active power and the generator bearing temperature. By combining time-series with a statistical distribution method, the characteristics of temperature changes are examined. The analysis results demonstrate that the monotonic and linear relationships between bearing temperature and individual operating parameters are weak, with all Spearman correlation coefficients below 0.4 and R² values below 0.01. To enhance fault identification accuracy, a multivariable nonlinear model based on Extreme Gradient Boosting (XGBoost) is developed to extract key features related to generator bearing temperature and identify potential faults. Verification through practical application cases indicates that the model can issue alerts approximately 20 minutes before a fault occurrs, exhibiting a strong fault identification ability. The findings provide effective decision-making support for the intelligent operation and maintenance of wind turbine generators.