Fault Detection and Classification in Active Distribution Network using Voltage and Current Features

This paper presents a fault detection and classification method in an active distribution network using relay voltage and current signals. The features are obtained using the empirical wavelet transform as well as other statistical indices, which are utilized to develop the machine learning-based protection models. Further, principal component analysis is deployed for the purpose of feature dimensionality reduction. The IEEE 13-bus network integrated with inverter-based distributed generation units modelled in PSCAD/EMTDC is considered in the study. A large data set comprising of diverse fault and non-fault events is obtained from the simulated network to obtain the protection models. Performances of the selected relays in the network are demonstrated regarding fault detection and faulty phase identification. The results exhibit robust and reliable response to the network events and demonstrate the efficacy of the proposed work particularly with regard to the uncertainty in the network operation scenarios. Moreover, the intelligent models can be implemented in the protection system along with the available protection schemes.