CEEMDAN-SP-CNN-based offshore wind turbine generator fault diagnosis method

The invention discloses an offshore wind turbine generator fault diagnosis method based on CEEMDAN-SP-CNN, and the method comprises the steps: obtaining multi-source time series data of offshore wind turbine generator state monitoring, carrying out the preprocessing, and constructing an original data set containing two-stage fault classification labels; performing modal decomposition on the feature signals in the original data set by adopting a CEEMDAN algorithm to obtain a plurality of intrinsic mode components IMF and a residual component, calculating the sample entropy of each IMF, and reconstructing the signals to form a one-dimensional time sequence feature data set; a one-dimensional time sequence in the one-dimensional time sequence feature data set is converted into a two-dimensional image by adopting a GAF, and a two-dimensional image feature data set is formed; and respectively inputting data in the one-dimensional time sequence feature data set and the two-dimensional image feature data set into th