According to transformer fault diagnosis for typical small sample data, Sthis paper proposes a support vector machine of power transformer fault diagnosis method (SVM) based on differential evolution algorithm optimization. This method use the differential evolution algorithm to optimize parameters C,g in support vector machine, using the optimized support vector machine (SVM) to fault diagnosis data of small sample, and comparing with other optimization algorithm of support vector machine (SVM).Experimental data shows that accuracy with this method higher than the grid search optimization method and particle swarm optimization algorithm ,and is very suitable for the fault diagnosis of power transformer.