Stochastic Characterization of Voltage Sag Occurrence Based on Field Data

Abstract

Computational methods for predicting voltage sags consider their occurrence as a Poisson process, although without confirmation from monitoring data, until now. In this paper, the stochastic nature of voltage sags is analyzed and discussed using monitoring data from 60 sites of the Portuguese Transmission Network, covering the years 2011–2015 (a total of 17 157 recorded voltage sags). A mathematical model to describe the voltage sag occurrence as a stochastic process is presented. The assumption of constant failure rates of the network elements implies that the occurrence of voltage sags is a Poisson process. However, that assumption is not valid if voltage sag clusters are included, as these imply considering time-dependent failure rates. Then, the time between voltage sags is described by an exponential distribution, if clusters are not included, and may be described by the gamma distribution, if including clusters. The boundaries of the adequacy of exponential and gamma distributions are assessed, based on monitoring data. The time of occurrence of monitored voltage sags are analyzed and results confirm that the Poisson process describes the occurrence of voltage sags when voltage sag clusters are disregarded. The gamma distribution fitting is also confirmed when clusters are included in the analysis.