Transformer Fault Diagnosis Based on Improving Kernel-Based Extreme Learning Machine

Abstract

The dissolved gas analysis (DGA) is the most commonly used method in field power transformer fault diagnosis, but its diagnosis result is unreliable. In order to further improve the accuracy of power transformer fault diagnosis, this paper analyzes the advantages and disadvantages of various classification and optimization algorithms, and finally presents a transformer fault diagnosis method based on cloud model and Quantum-behaved Particle Swarm Optimization(QPSO) optimization Kernel-based Extreme Learning Machine(KELM). The method applies the kernel function to the Extreme Learning Machine, and maps the lowdimensional nonlinear relation to the high-dimensional linear space, which avoids the dimension disaster and reduces the calculation cost. It has further improved the fault diagnosis ability, For the simultaneous parameter problem, the cloud model is used to optimize the shrinkage-expansion factor of the quantum particle swarm. The parameters of the kernel limit learning machine are optimized by the combination of the cloud model and the quantum particle group. And then the method has the advantages of strong global search ability, high searching precision and fast convergence speed. Finally, the validity of the proposed method is proved by comparing with other fault diagnosis methods.