A PCA-based approach for substation clustering for voltage sag studies in the Brazilian new energy context

Abstract

Voltage Sags are the most disturbing power quality deviation to the sensitive industrial loads causing production losses and other impacts to the equipment end users and distributors of electricity. A new methodology for estimation of voltage sag patterns and clustering of distribution substations with similar features for voltage sags is the main contribution of this paper. The focus is on the regulation of this phenomenon in Brazil. Network modeling and faults simulations in transmission and distribution levels were performed in a Brazilian actual distribution system with 17 substations in order to get information regarding the number of voltage sags caused in the substations bus bars. Principal Component Analysis was applied on a significant number of variables containing relevant information of power quality and design features of the substations and storing the major Principal Components Scores. Furthermore, Clusters of substations with similar characteristics to voltage sags were formed and assigned a specific Membership to each substation. The expected annual number of voltage sags is the solution. A 95% Confidence Interval for the total number of voltage sags was estimated, leading to the classification of the main variables related to the voltage sags by clusters formed. Clusters of Substations with similarities for regulatory purpose of voltage sags were the principal results, followed by classification of the main variables associated with voltage sags by the clusters formed. A critical analysis of the results and a comparison among different Clustering Methods reaffirmed the assertiveness of the proposed methodology for the management of power quality associated with voltage sags, by the Regulator and the distribution utilities in Brazil.