Machine Learning-Based Backup Protection for Wind Farm Integrated Transmission System

The primary distance relay protects up to 80% of line length. However, backup protection is provided for the next section of line in zone-II and zone-III of the distance relay. If the backup protection fails, the system may observe cascaded tripping. This paper proposes a backup protection scheme for fault classification and location identification in a wind farm-integrated transmission network. For this reason, the IEEE 14 bus system is modified and modeled in PSCAD/EMTDC software and simulated for faults at various locations for collecting high-resolution voltage measurements. Voltage measurements are utilized for determining the discrete wavelet transform (DWT). The DWT coefficients are used for feature extraction and feature selection for fault classification and location identification. A comparison between the accuracy and speed of two popular classification approaches support vector machine, and k-nearest neighbors has been made for fault type and location identification.