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

With the increase in penetration rate of distributed generation under dual-carbon goals, hazards of unplanned islanding caused by grid-connection of distributed generation systems have attracted attention. Due to the dilution effect of characteristic parameters in multi-machine grid-connected systems, active and passive hybrid islanding detection is prone to have islanding detection blind zones. Therefore, an islanding detection method suitable for complex multi-machine grid-connected systems of distributed generation is proposed. To verify the feasibility of the proposed islanding detection method, a simulation model of the distributed generation grid-connected system was built in MATLAB/Simulink based on IEC 61850-7-420 standard. Two different operating conditions of islanding and grid-connection, were simulated to obtain voltage and current waveforms of each node, which were converted into time series. Then, time-frequency spectrum analysis and calculation were performed based on Short-Time Fourier Transform (STFT) to generate time-frequency spectrum. After feature extraction, training iterations were carried out using Convolutional Neural Network (CNN) algorithm to obtain the optimal islanding detection model. Simulation results show that the proposed method achieves a detection accuracy of 99.84% on the independent test set, with a missed detection rate of 3.4% under five-fold cross-validation, which significantly reduces the non-detection zone.