Islanding and fault detection of inverter based distributed generations using wavelet packet transform and ensemble

• Published in: Electric power systems research 2024-06, Vol.231, p.110356, Article 110356
• Main Authors: M.A., Ajith, R.M., Shereef
• Format: Article
• Language: English
• Subjects: Distributed generation; Ensemble; Fault detection; Feature extraction; Islanding detection; Microgrids; Wavelet packet transform
• ISSN: 0378-7796; 1873-2046
• DOI: 10.1016/j.epsr.2024.110356

The adoption of Distributed Generation (DG) technology has seen a substantial level of advancement in recent years. This has created potential protection issues in microgrids during faults and islanding. This paper proposes a fault and islanding detection method using wavelet packet transform (WPT) and Ensemble. A new index called Normalized RMS Node power index (NRNPI) is defined based on WPT to differentiate between transient behavior during faults and islanding, and the ensemble acts as a classifier. The ability of the method in detecting islanding and both grounded and non-grounded faults is demonstrated through simulations of a test system in Matlab Simulink. Real time simulation of the proposed method is done in OPAL-RT using UDP-IP communication protocol. The proposed method as an islanding detection technique gives an accuracy of 100% and 99.32 % as a fault detection technique under the test conditions. Results of both simulations and Hardware in Loop (HIL) implementation is promising. The proposed method has the potential of extending the idea for other transient analysis in the microgrid. •Method capable of detecting islanding and faults in AC microgrid.•Wavelet packet transform used for feature extraction from measured signals.•Ensemble machine learning technique is used for classification of faults and island.•Applicable for microgrids with single distributed generator.•Fast and accurate islanding and fault detection method.