Positive Sequence Currents Similarity Based Active Distribution Network Protection Scheme Using One-Class Classification

Overcurrent protection scheme used widely in distribution networks cannot meet the selectivity and reliability requirements of active distribution networks. Besides, due to the complex fault characteristics of distributed generators, the new schemes based on single fault characteristics have not made breakthroughs. This paper proposes a new protection scheme based on the similarity of positive sequence currents (PSCs) at both ends of the line, which utilizes a one-class classification algorithm to fuse multiple fault characteristics of PSCs. Considering the data scarcity of internal and external faults, the normalized PSCs of normal operation are used to reflect the characteristics of external fault because the normalized PSCs of them have nearly same similarity characteristics theoretically but distinctly differ from internal fault, where the PSCs are integrated as a feature vector and normalized by maximum value normalization method. Besides, the phase angle difference of PSCs, caused by the data asynchronous problem, is corrected by calibrating the sudden change time of PSCs for sufficiently utilizing phasor angle beside amplitude. In summary, the normal operation data is firstly normalized and corrected with a certain correction error, then serves as input to train an OCC algorithm. The determined classification boundary is used for protection criterion setting to discriminate external fault from internal fault. The PSCAD simulations verify the effectiveness of the proposed protection scheme.