LSTM Recurrent Neural Network Classifier for High Impedance Fault Detection in Solar PV Integrated Power System

This paper presents the detection of High Impedance Fault (HIF) in solar Photovoltaic (PV) integrated power system using recurrent neural network-based Long Short-Term Memory (LSTM) approach. For study this, an IEEE 13-bus was modeled MATLAB/Simulink environment to integrate 300 kW PV systems for analysis. Initially, three-phase current signal during non-faulty (regular operation, capacitor switching, load transformer inrush current) and faulty (HIF, symmetrical unsymmetrical fault) conditions were used extraction features. The processing technique Discrete Wavelet Transform with db4 mother wavelet applied extract each phase's energy value features training testing classifiers. proposed LSTM classifier gives overall classification accuracy 91.21% a success rate 92.42 % identifying HIF network. prediction results obtained from proffered method are compared other well-known classifiers K-Nearest neighbor's network, Support vector machine, J48 based decision tree, Naïve Bayes Further, classifier's robustness is validated by evaluating performance indices (PI) kappa statistic, precision, recall, F-measure. reveal that network significantly outperforms all PI techniques.