In order to improve the fault diagnosis accuracy of transformer

we propose a method which is based on least squares twin support vector machine（LS-TSVM） optimized by bat algorithm（BA）. To solve the multi-classification problem of transformer fault diagnosis in this method

a Huffman tree is established from the bottom to up by calculating the inter class dissimilarity matrix

and a multi-classification fault diagnosis model is built based on the LS-TSVM

then a bat algorithm is used to optimize the parameters of the LS-TSVM classifier in the model. The simulation results of using this method to diagnose the fault of a transformer show that

compared with particle swarm optimization-support vector machine（PSO-SVM）

the proposed method not only shortens the training time

but also achieves a higher accuracy. The diagnostic accuracy for high-temperature overheating and low-energy discharge faults is obviously higher than that of PSO-SVM. The simulation results show that the method has high superiority in transformer fault diagnosis.