Análise de Componentes Principais para diagnóstico pós-falta em sistemas elétricos de grande porte

Abstract

Synchronized phasor measurement systems are being widely used around the world, and have become essential elements in the evolution of the operation of large electrical systems. These systems are based on phasor measurement units, called PMUs, which are capable of recording data with a high sampling rate, thus generating a huge dataset. This work presents a methodology for selecting data for post- fault analysis, through dimensionality reduction, using principal component analysis. To validate the proposed methodology, real data related to a recent occurrence in the national interconnected system were used. The validation of the obtained results was done using an anomaly detection algorithm. With the application of the methodology, the possibility of using a data set smaller than the original was proved, while maintaining the characteristics of the attributes. This automatic data selection methodology benefits post-fault analysis due to the performance gain obtained by reducing the electrical attribute dataset, without significant loss in the accuracy of anomaly detection capability.