Abstract:
Overcurrent relays are widely used for power systems protection. Transmission side uses more directional type relays, while distribution systems, e.g., radial and ring-main subtransmission systems use non-directional types. The fault direction may be forward (between the relay and the grid), or reverse (between the relay and the source), the normal power flow being from the source to the grid. Traditional directional overcurrent relays utilize the reference voltage phasor for estimating the direction of the fault. This requires measurement of both current and voltage using the respective sensors. This makes the directional overcurrent relays prohibitively more costly than the non-directional type. In this paper, a novel current-only directional detection possibility is highlighted. This paper describes a neural network (NN)-based approach for current-only directional detection in an overcurrent relay. Intelligent techniques like NN prove to be computationally less intensive than methods like discrete Fourier transform, Kalman filtering, etc. for phase angle computation. Implementation of different types of NNs, experiments along with application results on fault direction classification are presented. Possible utilizations of the current-only directional relay in the distribution side automation are described, which is a key focus area for enabling the smart grid.
