Single-phasing detection and classification in distribution systems with a high penetration of distributed generation

Abstract

This paper presents a robust wavelet-ANN based algorithm for single-phasing detection and single-phasing classification in distribution systems with a high penetration of distributed generation (DG). In traditional vertically integrated distribution systems, single-phasing events are detected easily, as the current of one of the phases is lost completely, resulting in a significant current unbalance ratio. However, this is not the case for distribution systems with a high penetration level of DG units, as the backfeed current from the DG units will support the current in the lost phase, hence masking the single-phasing operation. In the proposed algorithm, the transient current signals generated at the onset of the single-phasing condition are combined into a modal signal. This signal is analyzed using discrete wavelet transform (DWT) to extract the feature vector denoting the distinctive features for each frequency band. Finally, artificial neural networks (ANNs) are used to detect single-phasing conditions and to classify the lost phase. Simulation results have confirmed the dependability and security of the proposed algorithm.