The effective monitoring and early warning of an offshore Wind Turbine (WT) gearbox operating state is of great significance to its operation and maintenance. Therefore, a Condition Monitoring (CM) method for an offshore WT gearbox based on Gated Recurrent Unit (GRU) and attention mechanism is proposed. In the training phase, the attention mechanism is introduced to automatically extract the correlation between the input features and the target modeling feature of an offshore WT Supervisory Control and Data Acquisition (SCADA) dataset. The GRU network is used to effectively extract the temporal relationships between the data, thereby establishing the normal behavior model of the WT gearbox. In the test phase, an Exponentially Weighted Moving-Average (EWMA) control chart is applied to monitor the output residuals between the predicted value and the true value of the target modeling feature, so as to realize real-time monitoring and early warning of the operating status of the gearbox. Finally, the validity and superiority of the proposed method are verified based on real data from the Donghai Bridge offshore wind farm. The results show that the proposed method can effectively monitor the gearbox state under both fault and normal operating conditions, and has higher accuracy and interpretability than the existing onshore monitoring method, and can reveal early failure trends earlier. This work is supported by the National Natural Science Foundation of China (No. 61873159) and the Project of Science and Technology Commission of Shanghai (No. 18020500700).