Neural networks aided analysis of harmonic effects on overcurrent relays

ABSTRACT In this thesis, harmonic effects on operational characteristics of static and electromechanical overcurrent relays are studied. As an experimental study, an inverse time induction disc type overcurrent relay, an inverse time static overcurrent relay and an instantaneous electromechanical overcurrent relay are considered. A dimmer circuit is used to represent nonlinear load. In the experimental study, the nonlinear load currents which have different harmonic spectrums are applied to overcurrent relay inputs. Afterward, pickup current and operating time values for inverse time over current relays as well as pickup current values for instantaneous overcurrent relays are measured. Harmonic spectrums of nonlinear load currents are obtained by using a harmonic analyzer. The data set which are obtained from experimental results are trained using Artificial Neural Networks and back propagation algorithm in order to simulate the current-time characteristics of the considered overcurrent relays. Therefore, the responses of relays for any nonsinusoidal current can approximately be defined by using the jtrained neural network. It is evident from experimental results that the increase of the total harmonic current distortion value of the current (THDj) causes operating current and time delay of the induction disc type inverse time electromechanical overcurrent relay to increase. However, the increase of the THDj value on the inverse time static overcurrent relay causes the operating current and time delay to decrease. As the total harmonic distortion increases, inverse-time static overcurrent relay nominal current and time-delaying increases. In conclusions, the presence of harmonics as an important factor in power quality, results in the increase of operating errors and unavailability of protection relays. Keywords: Harmonics, overcurrent relays, artificial neural networks. xn