Classification and Localization of Faults in AC Microgrids Through Discrete Wavelet Transform and Artificial Neural Networks

Abstract

The widespread integration of renewable energy sources to the main electrical grids has led to the increased adoption of AC microgrids. However, the protection of AC microgrids is a challenging task due to inverter interfaces, bidirectional power flow, multiple modes of operation and the requirement for selective phase tripping. This paper presents an innovative artificial neural network (ANN) based approach for fast and accurate identification and localization of symmetrical and asymmetrical faults occurring in the distribution networks of AC microgrids. In the proposed methodology, the three phase and the neutral currents which are sampled at either ends of the distribution lines, undergo discrete wavelet transform to extract the features exhibited during faults in the network. These features are used by two neural networks for classification and localization of the fault. To achieve high accuracy and computational efficiency, the network architectures of the ANNs are optimized, and the extracted features contain the detailed information required for ANNs to clearly distinguish different fault types and locations. A comprehensive evaluation and validation reveal that the proposed scheme accurately and efficiently classifies and localizes faults in AC microgrids. The existing research gap of fault localization in AC microgrids is also addressed through this approach.