Search

Fault detection and classification in microgrid becomes a steep challenging issue due to its inherent characteristics, such as operating in both grid-connected and islanded modes, the intermittent nature of distribution generations, and the variation of fault current magnitudes. The microgird often experiences a symmetrical fault and three types of unsymmetrical faults due to the natural causes such as heavy wind, lightning, aging of cables and/or poles, and others. Therefore, an accurate and precise fault diagnosis model is needed for proper and smooth operation of the microgrid. In this paper, a deep learning model based on Long Short-Term Memory is addressed for microgrid faults diagnosis. The Matlab/Simulink environment is used for both microgrid model implementation and faults simulation. The model is implemented based on modified IEEE 13 Node Test Feeder. The faults are simulated on a distribution line of the microgrid starting from the relay location for collecting the training and testing datasets which include ten different shunt faults. The proposed deep learning model can successfully identify faults in 1/4 cycle data window of three-phase voltages and currents for both modes of operation. The average testing classification accuracy and precision of the proposed model are 99.89 % and 0.9934 when the microgrid is operated in grid-connected mode and the average classification accuracy and precision are 99.88 % and 0.9930 for the islanded mode of operation. [ABSTRACT FROM AUTHOR]

This paper presents a failure model of heating filaments as used in semiconductor manufacturing tools for growing epitaxial films. Attempts were endeavoured to provide feasible physical insights on the observations in changes of impedance as indirectly derived via readings of voltage-current measurements, instead of conventional data-driven black-box approach of statistical regressions. The proposed model is able to predict the magnitudes of these impedance changes and the classification result lends itself to the setting up of quantifiable control limits to enable early warnings of fault detections in process controls. [ABSTRACT FROM PUBLISHER]

This paper develops a methodology for application in distribution network fault detection and classification. The proposed methodology is based on wavelet energy spectrum entropy decomposition of disturbance waveforms to extract characteristic features by using level-4 db4 wavelet coefficients. Thus, few input features are required for the implementation. Different simulation scenarios encompassing various fault types at several locations with different load angles, fault resistances, fault inception angles, and load switching are applied to the IEEE 34 Node Test Feeder. In particular, the effects of system changes were investigated by integrating various Distributed Generators (DGs) into the distribution feeder. Extensive studies, verification, and analysis made from the application of this technique validate the approach. Comparison with statistical methods based on standard deviation and mean absolute deviation has shown that the method based on log energy entropy is very reliable, accurate, and robust. [ABSTRACT FROM PUBLISHER]