Comparison and Validation of Statistical Methods for Predicting Power Outage Durations in the Event of Hurricanes

Abstract

This article compares statistical methods for modeling power outage durations during hurricanes and examines the predictive accuracy of these methods. Being able to make accurate predictions of power outage durations is valuable because the information can be used by utility companies to plan their restoration efforts more efficiently. This information can also help inform customers and public agencies of the expected outage times, enabling better collective response planning, and coordination of restoration efforts for other critical infrastructures that depend on electricity. In the long run, outage duration estimates for future storm scenarios may help utilities and public agencies better allocate risk management resources to balance the disruption from hurricanes with the cost of hardening power systems. We compare the out-of-sample predictive accuracy of five distinct statistical models for estimating power outage duration times caused by Hurricane Ivan in 2004. The methods compared include both regression models (accelerated failure time (AFT) and Cox proportional hazard models (Cox PH)) and data mining techniques (regression trees, Bayesian additive regression trees (BART), and multivariate additive regression splines). We then validate our models against two other hurricanes. Our results indicate that BART yields the best prediction accuracy and that it is possible to predict outage durations with reasonable accuracy.

Predicting Urban Reservoir Levels Using Statistical Learning Techniques

“…Supervised learning theory is a branch of statistical learning methods that has been extensively applied to areas ranging from risk and resilience analysis to hydrological modeling 7–10 . Supervised learning models vary widely in their degree of complexity, stability, flexibility and interpretability, and can be categorized as parametric, semi-parametric or non-parametric methods.…”

Proactive Recovery of Electric Power Assets for Resiliency Enhancement

“…The time to repair for each damaged component is stochastic in its nature. In [27], it was shown that the repair time is not only a function of the number of crews, but also other factors, e.g., geographical characteristics of the area that limit crews access can affect the repair time. On the other hand, because of the variation in skill level of the repair crew along with the random nature of the degree of damage, the time to repair need to be considered as a random variable.…”

“…However, the availability of past data could be a challenging issue which limits the model practicality. Reference [27] compared the regression methods and data mining techniques for predicting power outage durations during hurricanes. The accuracy of Bayesian additive regression trees (BART) outperformed the other models in their study.…”

Post-disaster infrastructure restoration: A comparison of events for future planning

“…While Nateghi et al [49] and Guikema et al [29] follow a similar approach, both estimating Contents lists available at ScienceDirect journal homepage: www.elsevier.com/locate/ijdrr outages as a function of climatic, topographical, land cover, human and material resources, and population densities amongst others. Ultimately these approaches require a significant amount of potentially difficult to obtain input data which can be limited in accuracy -if accessible at all.…”