Automated classification of power quality disturbances using RBF and SVM neural networks

This paper presents classification results of different power quality disturbances. SVM and RBF neural networks are considered as appropriate classifiers for power quality issues, however SVM networks show better performance. Simulation of disturbed signals by parametric equations enabled the assessment of signal parameters influence on classification rate. Positive results encouraged further research. Model of supply system suffering from sags was simulated. Independent from line length and sag duration the classifier was set to recognize different sag types. The idea of space phasor was applied to obtain distinctive patterns from three phase system. Wavelet transform was used to find the beginning of sags. Positive classification results were obtained.