An improved ANN based admittance relay using pre-processed inputs

This paper addresses some of the issues associated with the conventional relay designs and presents an improved distance relaying pre-processing algorithm. Instantaneous current and voltage values obtained directly from the power system have been used to obtain the processed inputs. The presented algorithm has been combined with a neural network approach to eliminate the process of phasor estimation, which is usually used in most numerical relaying algorithms. The neural network has been trained to recognize the phase difference between the processed inputs, and therefore eliminates the need of calculating phasors. The processed inputs given to the neural network have a direct relationship with the outputs expected from a relay, which helps to use a data window lesser than one full cycle to accurately detect faults, making the algorithm faster than traditional designs. The neural network based relay has been trained using pure sinusoidal values and tested on a 17-bus power system simulated in PSCADtrade. The results show that the relay is able to detect faults in lesser time as compared to conventional relay algorithms while maintaining the integrity of relay boundaries.