Converter transformer is the key equipment for power conversion and transmission in HVDC transmission system

and its operation reliability is very important to the overall stability of power grid. In this paper

the structural optimization and fault diagnosis of converter transformer windings are systematically studied

and a complete method system of multi-field coupling modeling+structural parameter optimization+winding fault diagnosis is proposed. Firstly

the simulation model of converter transformer with electromagnetic-structural force field coupling is constructed

and the variation law of vibration characteristics under harmonic excitation is analyzed. Secondly

the minimum vibration is taken as the optimization objective

and six key structural parameters

such as winding height

thickness

and spacing

are selected. The gradient boosting regression tree model combined with Bayesian optimization algorithm is used for optimization

so that the maximum vibration displacement is reduced by 11.93% compared with the initial design. Furthermore

a diagnosis method based on vibration signal is proposed for the faults that may still occur in the optimized winding. By using the method

the best position of the tank wall measuring point can be obtained through correlation analysis

the typical fault simulation model of the winding can be constructed

the fault diagnosis model can be established based on TST-XGBoost

and the fault recognition accuracy of 98.33% can be realized. The research results not only provide an effective method for the vibration reduction design of converter transformer

but also provide theoretical basis and technical support for winding state evaluation and intelligent diagnosis.