Abstract: Mine transformers are mainly used in mines containing flammable gases and coal so as to provide power for coal mining machines, transporter and lighting systems. At present, because the accuracy of mine transformer fault detection is not high, this paper presents a fault diagnosis method for mine transformers based on improved cuckoo algorithm and support vector machine (SVM). In this paper, the improved cuckoo algorithm is introduced to optimize the parameters of the SVM, and the SVM model with the best parameters can obtain, then the transformer faults are classified by using the support vector machine. Finally, the effectiveness of algorithm and detection method is verified in this paper. Matalab simulation results show that using algorithm and diagnosis model in this article provides better overall performance than other method.