Discrimination of Stator Single-Line-to-Ground Fault Considering Distribution Characteristics of the Data for Powerformer

Abstract

Some traditional protection schemes cannot detect a faulted Powerformer when a stator single-line-to-ground fault occurs in parallel Powerformers. A new discrimination method based on kernel principal component analysis (KPCA) and Fisher discrimination analysis considering the distribution characteristics of the data is proposed in this paper. First, the algorithms of the KPCA and the Fisher discrimination are introduced. In order to increase the correct rate of the fault recognition, a kernel function for the distribution characteristics of the data is selected according to the error rate under the historical fault data, which is different from the existing application validation methods. Then, the detected data in the system model are processed to detect whether a stator single-line-to-ground fault occurred in the Powerformer using the KPCA and the Fisher discrimination analysis based on an optimal kernel function. Finally, simulation results show that the proposed scheme can exactly detect the fault for a Powerformer in different neutral grounding systems when a stator single-line-to-ground fault occurred.