Transformer Intelligent Diagnosis Method Based on AFSA-MKELM Algorithm

Abstract

In the kernel extreme learning machine, a single global kernel function or a local kernel function can not balance the contradiction between fitting and generalization well, so replace the single kernel function with a mixed kernel function to improve the learning and generalization ability of the model. Based on this, a Multi-Kernel Extreme Learning Machine (MKELM) algorithm was proposed. Considering the parameter sensitivity characteristics of the MKELM, the artificial fish swarm algorithm (AFSA) was used to optimize the parameters. In summary, a hybrid algorithm was proposed and applied to the field of transformer fault diagnosis. Compared with the extreme learning machine algorithm of single kernel function and SVM algorithm, the diagnostic accuracy of this algorithm is higher.