At present，the mainstream operation and maintenance technology of electromechanical equipment in water conservancy projects has developed from automatic operation and maintenance to intelligent operation and maintenance.Using artificial intelligence to assist or even partially replace manual decision-making can further improve the quality and efficiency of water conservancy operation and maintenance.A detailed fault diagnosis method based on deep learning and cosine similarity is proposed for the faults occurring in practical operation of equipment.Firstly，the deep learning method is used to identify and classify the current fault.On this basis，according to the classification results，the cosine similarity algorithm based on TF-IDF is used to further refine the diagnosis through the similarity matching of fault phenomenon text.Finally，the follow-up operation and maintenance work is carried out according to the fault refinement diagnosis results.Taking transformer fault as an example，the experimental results show that this method has good diagnostic performance and is suitable for popularization and application.