Failure interruption often causes large blackouts in power grids, severely impacting critical functions. Because of the randomness of power failure, it is difficult to predict the leading causes of failure. ASAI, an essential indicator of power-supply reliability, can be measured from the outage time series. The series is non-stationary stochastic, which causes some difficulty in analyzing power-supply reliability. Considering that the time series has long-range dependence (LRD) and self-similarity, this paper proposes the generalized Cauchy (GC) process for the prediction. The case study shows that the proposed model can predict reliability with a max absolute percentage error of 8.28%. Grey relational analysis (GRA) has proved to be an effective method for the degree of correlation between different indicators. Therefore, we propose the method, which combines both GC and GRA to obtain the correlation coefficients between different factors and ASAI and to get the main factors based on this coefficient. The case study illustrates the feasibility of this approach, which power enterprises can employ to predict power-supply reliability and its influencing factors and help them identify weaknesses in the grid to inform employees to take protective measures in advance.
Keywords: failure interruption; generalized Cauchy process; grey relational coefficient; long-range dependence; reliability of distribution systems.