A novel principle of frequency characteristics-based current differential protection for outgoing lines of wind farms

• A frequency component analysis method for differential current in wind farm is proposed. • A transient differential protection scheme within a 5 ms time window is proposed. • The analysis result and protection scheme are verified through actual physical dynamic experiment. With the development of renewable energy sources, conventional power frequency-based protection schemes face performance decrease problems. Studying the transient characteristics of faults is crucial for increasing the protection performance for outgoing lines of wind farms. In this paper, the transient dominant frequency component of the fault differential current in the outgoing line is quantitatively analyzed based on the lumped parameter line equivalent method and the frequency impedance of the wind farm. At the initial stage of fault transient, a novel principle of current differential protection within a 5 ms fault window is proposed using the composition difference between the dominant frequency component and power frequency component in external and internal faults. Afterwards, based on the time–frequency analysis method of wavelet transform, the power frequency component and the transient dominant frequency component are calculated and considered as the action and braking energies, respectively. The composition of the frequency band for energy calculation is then optimized based on the characteristics of the wind farm outgoing line. This allows to significantly improve the sensitivity and reliability of the protection. Finally, the proposed scheme is evaluated through PSCAD simulation and dynamic experiment. The results of the dynamic experiment, which relies on physical models, demonstrated the correctness of the theoretical analysis and the high performance of the proposed protection scheme. The experiment results show that the proposed protection scheme, which is not affected by the CT saturation and requires sampling frequency of only 10 kHz, can identify internal faults in 6–7 ms.