Smart grids power quality analysis based in classification techniques and higher-order statistics: Proposal for photovoltaic systems

This paper compares classification techniques to identify several power quality disturbances in a frame of smart metering design for smart grids with high penetration of PV systems. These techniques are: Linear Discriminant Analysis (LDA), Nearest Neighbor Method (kNN), Learning Vector Quantization (LVQ) and Support Vector Machine (SVM). For this purpose, fourteen power-quality features based in higher-order statistics are used to assist classification. Special attention is paid to the spectral kurtosis, whose nature enables measurement options related to the impulsiveness of the power quality events. The best technique of those compared is selected according to correlation and mistake rates. Results clearly reveal the potential capability of the methodology in classifying the single disturbances. Concretely, the SVM classifier obtained an average correlation rate of 99%. Hence, concluding that it is a robust classification method.