Hurricanes have caused severe damage to the electric power distribution system in the U.S., leading to widespread, prolonged power outages. Electric power is critical to post-hurricane disaster response as well as to long-term recovery for impacted areas, and power restoration depends on effective pre-storm planning and resource allocation. Power outage estimates form the basis of pre-storm decisions that utility companies must make about the number of crews to request through mutual aid agreements and the locations where crews and materials should be staged in preparation for a recovery effort. Managing power outage risk and properly preparing for post-storm recovery efforts requires rigorous methods for estimating the number and location of power outages before a storm makes landfall. These estimates must be geographically detailed and accurate while also accounting for the complicated relationships between a number of possible explanatory variables and power outages. Previous research in estimating power outages during hurricanes has relied on negative binomial generalized linear models (GLMs) that use the standard linear relationship between the log of the expectation of the response variable Y (i.e., the number of power outages in a geographic area) and a set of explanatory variables given in the matrix X . However, these models have been shown to substantially overestimate the number of power outages in urban areas in some cases (Han et al. 2007a), and they do not provide insight into the possibility of non-linear relationships between the elements of X and the response variable. This paper shows how generalized additive models (GAMs), a class of semi-parametric regression models, can be used to more accurately estimate power outages during hurricanes. This is done through an analysis of a data set consisting of power outages in 6,681 3.66 km (12,000 foot) by 2.44 km (8,000 foot) grid cells during five hurricanes in the service area of a large, investor-owned utility company serving a large portion of a Gulf Coast state that has been impacted by a number of hurricanes in the past. The results show that GAMs can provide more accurate predictions of the number of power outages in each geographic area of a utility company's service area and a better understanding of the response of the system than GLMs do. The explanatory variables used in the regression model include information about (1) the winds experienced in each grid cell during each hurricane, (2) the long-term precipitation and the soil moisture levels in each grid cell at the time of the hurricane, (3) the power system components in each grid cell, and (4) land use and land cover in each grid cell. The underlying goal in developing an outage count prediction model is to provide a basis for choosing how many crews and materials to request from other utility companies as a hurricane is approaching and where to position these crews and materials in order to restore electric power as quickly as possible without incurring unnecessary expense.
[1]
Haibin Liu,et al.
Spatial generalized linear mixed models of electric power outages due to hurricanes and ice storms
,
2008,
Reliab. Eng. Syst. Saf..
[2]
Samuel T. Ariaratnam,et al.
Assessment of Infrastructure Inspection Needs Using Logistic Models
,
2001
.
[3]
Jery R. Stedinger,et al.
Negative Binomial Regression of Electric Power Outages in Hurricanes
,
2005
.
[4]
Balvant Rajani,et al.
Comprehensive review of structural deterioration of water mains: statistical models
,
2001
.
[5]
Seth D. Guikema,et al.
Estimating the spatial distribution of power outages during hurricanes in the Gulf coast region
,
2009,
Reliab. Eng. Syst. Saf..
[6]
S. Wood.
Generalized Additive Models: An Introduction with R
,
2006
.
[7]
S. Guikema,et al.
Statistical models of the effects of tree trimming on power system outages
,
2006,
IEEE Transactions on Power Delivery.
[8]
R. Burnett,et al.
The Effect of Concurvity in Generalized Additive Models Linking Mortality to Ambient Particulate Matter
,
2003,
Epidemiology.
[9]
Francesca Dominici,et al.
New problems for an old design: time series analyses of air pollution and health.
,
2003,
Epidemiology.