Aiming at the problems of low accuracy and long running time of support vector machines(SVM) transformer fault diagnosis model, a transformer fault diagnosis model based on bald eagle search algorithm (BES) is proposed. Firstly, four test functions are selected to test the performance of BES algorithm, and compared with cuckoo algorithm (CS), artificial bee colony algorithm (ABC) and firefly algorithm (FA). The results show that BES algorithm has better optimization performance both in convergence speed and generalization ability. Then, the BES algorithm is used to optimize the kernel function parameters g and c of SVM, and the BES-SVM transformer fault diagnosis model based on dissolved gas analysis (DGA) in oil is established. The simulation experiments are compared with ELM, SVM, CS-SVM, ABC-SVM, FA-SVM diagnosis models. The results show that the comprehensive accuracy of the BES-SVM model is 98.67%, which is 22.67%, 20%, 13.34%, 12%, 10.67% higher than the above comparison fault diagnosis models, and the running time is the shortest. The proposed BES-SVM transformer fault diagnosis model has better fault diagnosis effect.