In order to solve the problems that the construction and fault diagnosis of bearing performance deterioration index are highly dependent on expert experience,many constraints and single practical application scenarios,a method for the construction and fault diagnosis of wind turbine bearing deterioration index was proposed,which combined the pelican optimization algorithm (POA),variational mode decomposition (VMD),and autoencoder.Firstly,the POA-VMD algorithm was used to decompose the vibration signals of the entire lifespan of the bearing into K intrinsic mode functions (IMF) by the adaptive approach,and K individual autoencoders were constructed for each IMF to capture their distinctive characteristics.Then the autoencoders were trained with the decomposed results of the normal vibration signals as the training sample,and bearing degradation index was constructed based on the output result of the model after the training was completed,and the early weak failure of the bearing was monitored with the help of the deterioration index.Finally,the envelope spectrum analysis of the IMF component reconstruction results of the vibration signal at the time of fault was carried out to determine the fault type.Experimental results validate that this method can not only clearly show the deterioration process of the bearing,but also have a high sensitivity to early weak faults,and can accurately diagnose the type of fault after the fault occurs.