The power transformer is the key equipment in electrical power system,once it breaks down,definitely will bring the huge economic loss.We first analyze transformer fault diagnosis through Fuzzy C means,immune algorithm,chaos optimization algorithm,then propose an improved immune Fuzzy clustering algorithm.In the algorithm,the fault samples of gray correlation degree and Fuzzy clustering objective function as parameters of the individual affinity function,we take Cauchy mutation in chaotic mutation operator,which makes the large majority of the probability of the search space as possible,resulting in greater variation,easily jump out local optima and maintain the population diversity.Experimental results show that the proposed algorithm applied to transformer fault diagnosis accuracy rate to more than 90%,the algorithm not only effectively overcome the traditional FCM clustering algorithm shortcomings easy to fall into local minimum,but also effectively suppresses the immune evolutionary process produces the "degradation" phenomenon.