To solve the problem of low recognition accuracy of transformer insulation oil gas fault diagnosis

the slime mold algorithm(SMA) is improved by the reverse learning strategy to form the improved slime mold algorithm(ISMA)

thus to improve the global optimization ability and optimize the support vector machine(SVM).An ISMA-SVM optimized fault diagnosis model is established

and the sample set is used for learning and training.The diagnosis and recoginition results are compared with that of the greywolf algorithm(GWO-SVM) and the particle swarm optimization(PSO-SVM)

it shows that the accuracy of the ISMA-SVM fault diagnosis and recognition is 93.3%

which is 6.66 and 10.66 percentage points higher than that of the GWO-SVM and PSO-SVM

respectively.