Machine Learning-Based Modification of Accelerated TRIP Schemes Deployed With Distance Relays

Faults which are in close vicinity from the ends of the transmission line are cleared with prolonged delays due to pessimistic setting of distance relays. A communication system based Accelerated Trip (AT) signal is used to clear such faults conventionally. However, protection schemes dependance on communication system to issue AT signal pose challenges in reliable operation. The implementation of such scheme is compromised in events of delay in communication signal transfer (channel latency), and failure of the communication systems referred as Loss of Signal (LoS). Currently, a backup communication system is used in such events. Researchers have extensively explored back-up protection schemes for AT signal issue based on local measurements to avoid dependance on the communication system. However, the robustness of the available methods is limited in dynamic power system events, and remote end breaker stuck condition. Citing the same, the authors propose a Machine Learning (ML) based method to correctly issue AT signal in dynamic power system events such as variation in fault resistance, power swings, and breaker arc re-striking. The ML based algorithm aims to classify the fault in internal, or external fault type category. Moreover. The proposed method can aptly issue AT signal even in breaker stuck condition. The efficacy of the proposed method is depicted using simulations in PSCAD. The proposed algorithm is validated with a scaled-down hardware setup of the transmission line section. Additionally, a COMTRADE (Common format for Transient Data Exchange) is generated using cases reported in public domain and the proposed method is also validated with the same.