With the continuous promotion of the dual-carbon strategy, the importance of renewable energy generation is becoming more and more important. Offshore wind turbines are located in remote areas, and maintenance after failure is more difficult. Therefore, the research on wind turbine fault prediction and identification technology is crucial. A machine learning framework based on Python sklearn and a fault warning method based on TensorFlow ，multiple feature extraction methods is adopted and convolutional neural networks is used for feature fusion and classification, which can achieve fault diagnosis and warning for offshore wind power systems. The design ideas, experimental steps, and experimental results of the method was introduced in detail, and this method was evaluated and analyzed.