As considering the complex operational conditions of offshore wind turbines, classifying the operational conditions can improve the accuracy of health condition evaluation of wind turbines and provide a reliable basis for the reasonable operation maintenance strategies. An operational conditions classification method based on kernel principal analysis (KPCA) optimized by particle swarm optimization algorithm (PSO) is proposed. The fitness function of kernel function parameter optimizations is constructed by considering within class scatter and between-class scatter of data to avoid the problem that choosing a proper kernel function parameter is difficult. The KPCA optimized by PSO is applied to data feature extraction and a classification model is built by using fuzzy C-means (FCM) clustering algorithm. The simulations on three groups data of UCI database prove the method''s validity. Finally, the method is used to classify the operational conditions of offshore wind turbines, and the results show that the proposed method can get much better classification effect than PCA+FCM and KPCA+FCM. It is helpful for establishing health condition evaluation model for each condition. This work is supported by National Natural Science Foundation of China (No. 51507098).