Wind turbine operation and maintenance (O&M) faces significant challenges due to the complexity of equipment, harsh operating environments, and the difficulty of real-time fault prediction. Traditional methods often fail to provide timely and accurate warnings, leading to increased downtime and maintenance costs. To address these issues, this study systematically explores an intelligent operation and maintenance method for wind turbines, utilizing digital twin technology and multi-source data fusion. Specifically, it proposes a remote intelligent operation and maintenance (O&M) framework for wind turbines based on digital twin technology. Furthermore, an algorithm model for multi-source operational data analysis of wind turbines is designed, leveraging a Whale Optimization Algorithm-optimized Temporal Convolutional Network with an Attention mechanism (WOA-TCN-Attention). The WOA is used to optimize the hyperparameters of the TCN-Attention model. Then, the gearbox fault alarm threshold and warning threshold are set using the statistical characteristics of the residual values, and the absolute value of the residuals is used to determine the abnormal operating state of the gearbox. Finally, the proposed method was validated using operational data from a wind farm in Xinjiang. With input data from multiple sources, including seven key parameters such as temperature, pressure, and power, the method was evaluated based on EMAE, ERMSE, and EMAPE. The results demonstrated that the proposed method achieved the smallest prediction error and provided effective early warnings 18 h and 33 min prior to actual failures, enabling real-time and efficient operation and maintenance management for wind turbines.