Hybrid islanding detection method using PMU‐ANN approach for inverter‐based distributed generation systems

• Published in: IET renewable power generation 2024-12, Vol.18 (S1), p.4453-4464
• Main Authors: Sarhan, Mohammad Abu, Barczentewicz, Szymon, Lerch, Tomasz
• Format: Article
• Language: English
• Subjects: DC‐AC power convertors; power distribution faults
• ISSN: 1752-1416; 1752-1424
• DOI: 10.1049/rpg2.13123

An essential component of guaranteeing the stability and safety of electricity distribution networks is islanding detection. In this work, a novel method for islanding detection which combined both phasor measurement units (PMU) and artificial neural network (ANN) is proposed. Using PMU measurements, the technique extracts features including phasor voltage, voltage frequency, and voltage rate of change of frequency (ROCOF), which later are fed into an ANN classifier. Using a huge dataset of more than a hundred thousand observations of both islanding and non‐islanding scenarios, testing was done on 24 distinct types of inverters in compliance with PN‐EN 62116 protocol criteria. The tests were carried out using Regenerative Grid Simulator Chroma 61815‐powered system which was connected in parallel to adjusting RLC load; the tested inverters were linked to a Photovoltaic Panels Simulator, the National Instruments cRIO‐9024 measuring equipment was used to carry out the measurements, MATLAB and LabVIEW were used for analyzing the data and results. With a testing accuracy of 99.05% and a training accuracy of 99.34%, the results demonstrate a high degree of accuracy. This work offers a practical solution for problems that occurred due to islanding phenomenon in power networks which can enhance the system dependability and security. A novel method for islanding detection which combined both phasor measurement units (PMU) and artificial neural network (ANN) is proposed. Using PMU measurements, the technique extracts features including phasor voltage, voltage frequency, and voltage rate of change of frequency (ROCOF), which later are fed into an ANN classifier.