Abstract: in the process of diagnosing faults in the planetary gearbox of wind turbines, the relationship between fault
characteristics and fault states is not unique, the effect of fault diagnosis directly using the output signal of planetary gearbox
cannot be guaranteed. A research method on fault diagnosis of wind turbine planetary gearbox based on knowledge graph
and GPT model was proposed. Using MOEA / D algorithm to decompose the fault feature selection problem into single
objective sub problems, combined with differential evolution strategy, select the feature with the highest correlation with the
fault. Using the selected features as input parameters for the GPT model, output a knowledge graph of planetary gearbox
faults in wind turbines under a small sample learning mechanism. Based on the mapping results of the planetary gearbox
operation data in the knowledge graph, determine the specific fault state of the wind turbine planetary gearbox. The test
results demonstrated that the method can effectively diagnosed various fault states, achieved a diagnostic accuracy of 0. 98,
and significantly improved the diagnostic performance.