Fault Section Identification in Distribution Networks with DFIG and PMSG Generators Using Current Transients

This paper presents a methodology for fault section identification (FSI) in distribution networks with embedded wind power generation. The phase currents are measured only at the distribution substation (DS), using a waveform window of two cycles (one before and one after the fault detection). The proposed approach is divided into two stages: the first stage, Fault Identification (FI), aims to identify whether a short-circuit fault lies on a main feeder or one of the branches effectively addressing the challenge of multiple fault locations that may arise when several branches correspond to the estimated fault point; the second stage, Fault Location (FL), estimates the distance between the DS and the fault location. The algorithm employs discrete wavelet transform (DWT) in combination with artificial neural networks (ANNs). Energy and Relative Energy Entropy, both in per unit (EPU and REEPU), are proposed and calculated from DWT decomposition, with regularization indexes applied to EPU and REEPU. These indexes serve as input to multi-layer ANN models, which work as classifiers for FI and predictors for FL. Various fault scenarios with different fault inception angle, fault type, fault resistance and fault location are simulated using MATLAB software and the IEEE 34-node benchmark feeder as test system. The results demonstrate that the proposed methodology performs effectively the FSI task, achieving an accuracy of up to 95% for FI and a maximum error of 5.2% for FL.