Discriminant-Analysis-Based Single-Phase Earth Fault Protection Using Improved PCA in Distribution Systems

This paper presents a new approach for single phase-earth fault protection in distribution systems. Traditional protection schemes are analyzed and compared with one another for effective fault feature extraction. The maximum-likelihood method is also carried out on several copula functions to find the optimal copula to fit the fault feature data. Then, copula rank correlation is calculated by the optimal copula, and the principal component analysis technique is improved to preprocess and reduce the dimension of the fault data by decomposing the copula rank correlation matrix instead of computing the eigenvalues and their eigenvectors of the covariance matrix. Finally, the distance discriminant function is defined and operation criterion is proposed, and distance discriminant analysis is used to discriminate the faulty feeder. Simulation results of a practical 10-kV distribution system show that the proposed approach of fault protection is able to achieve better identification accuracy than the hierarchical clustering algorithm and Fuzzy c-means algorithms-based approaches.