Research on an islanding detection method suitable for distributed generation grid-connection complex system

Description

Wen Sun, Sihan Yu, and Zhengye Jiang propose a way to detect islanding in distributed power generation, in a 2026 engineering study. Islanding happens when a generator keeps feeding loads after disconnecting from the grid, a serious fault that protection systems must catch quickly. In systems with generators, the signals dilute, and standard detectors develop blind zones where they miss the event.

The team built a simulation of a grid-connected distributed generation system in MATLAB and Simulink, under the IEC 61850-7-420 standard. They recorded voltage and current waveforms at nodes under islanded and grid-connected operation. Those waveforms became time series, time-frequency spectrograms through a short-time Fourier transform. A convolutional neural network trained on spectrogram features learned to classify system state. On a held-out test set the method reached an accuracy of 99.84 percent, and the authors report a much shrunken non-detection zone compared with standard approaches.

Catalog record

Source: https://europepmc.org/articles/PMC13218529