Machine Learning-based Modification of Communication Assisted TRIP Schemes deployed with Distance Relays

Abstract

Inadvertent over/ under reach of the distance relays due to variable DC offset, fault resistance, infeed, and CT and PT errors cause the pessimistic setting of the distance relays. Hence the faults occurring close to either end of the transmission section under protection are cleared with unintentional time delays. To avoid the same, distance relays are incorporated with communication-assisted TRIP schemes where the relay issues a TRIP signal in addition to the distance relay TRIP. The communication-assisted TRIP schemes are Direct/Permissive Over/Under Reach Trip schemes (DUTT/ DOTT/ PUTT/POTT). However, certain weaknesses in the communication-assisted TRIP schemes result in unintentionally long fault-clearing times and nuisance TRIP. In the presented study, the weaknesses of communication-assisted TRIP schemes are elaborated and a data-driven, adaptive Machine Learning (ML) based solution is proposed to improve the selectivity of the relaying scheme. Also, with the deployment of the ML-based relay, the fault sensitivity is improved. Moreover, an ensemble-based approach to data mining is presented that improves computational speed by reducing the computational burden.