Phase selective protection in microgrids using combined data mining and modal decomposition method

Abstract This paper presents a combined data mining and modal decomposition method for phase selective protection in microgrids. A comprehensive fault analysis is performed in time domain to generate two feature sets as the most representative features by using modal decompositions of Clarke and symmetrical components transforms. The generated feature sets are then used in two procedures. In the first procedure, the Clark-based feature set is input into a single classifier to classify faults. In the second procedure, two series combined classifiers are used, in which the first classifier identifies the fault type using symmetrical components-based feature set and the second classifier determines the faulted phase(s) using the Clarke-based feature set. The proposed method is evaluated by extensive simulation of the standard IEC microgrid in islanded and grid connected modes, with radial or meshed topologies under different fault conditions. The simulation study validates the method, in particular it shows that the second procedure with the series combined classifiers provides significantly higher accuracy in fault type classification. The robustness of the proposed method is tested against measurement errors and external faults. The proposed method obviates the need for feature selection or additional processing steps and provides high accuracy at lower sampling frequency.