Neural estimation of voltage-sag waveforms of non-monitored sensitive loads at monitored locations in distribution networks considering DGs

Concerns about power quality (PQ) in distribution networks have necessitated the use of PQ measuring/monitoring equipment to detect and analyse PQ problems. As the installation of such equipment at all locations is not economically feasible, minimum number and optimum locations of PQ monitors have been sought in recent researches to meet both economical efficiency and monitoring capability. Focusing on voltage sags, this paper attempts to further reduce the number of PQ monitors obtained by optimum allocation approaches while keeping the desired monitoring capability. Growing number of electric equipment with high sensitivity to voltage sags have raised more concerns about financial losses associated with voltage sags in comparison to other PQ problems. Here, a neural estimator placed at a monitored location is proposed to estimate instantaneous voltage-sag waveforms of a non-monitored sensitive load using local measurements. Echo state network (ESN) is used as the voltage-sag waveform estimator (VSWE). Because of increasing penetration of distributed generations (DGs) and their impacts on voltage-sag performance, they are considered to challenge the estimation task. The proposed ESN-based VSWE is examined on the IEEE 37-bus network. Tests for extensive unseen cases with different fault resistances, fault inception-angles, fault locations and fault types under different load demands and network conditions show acceptable high accuracy of estimations during fault and fault clearing sequences. Furthermore, the performance of the VSWE is investigated during transients due to switching capacitors, DGs, loads and lines. (C) 2012 Elsevier B.V. All rights reserved.