Modeling the Effects of Winter Storms on Power Infrastructure Systems in the Northern United States
Author: Jordan Vick Pino
Publisher:
Published: 2019
Total Pages:
ISBN-13:
DOWNLOAD EBOOKWinter storms cause significant damage to the power infrastructure system each year in the United States. These storms leave millions without power for an extended period of time, resulting in substantial economic losses. Utility companies seek to lower restoration costs and better prepare for such events. Therefore, many are investing in decision-support tools such as predictive models. These tools, coupled with industry experience, can aid in pre-storm planning and post-storm restoration activities, thus lowering costs and reducing outage extent and duration. This research developed winter storm impact models for FirstEnergy, an investor-owned utility company headquartered in the Midwestern United States.Specifically, three main objectives were addressed: (1) development and validation of a service territory winter storm impact model for FirstEnergy, (2) development and validation of regionally-defined winter storm impact models for FirstEnergy, and (3) a comparative analysis of the service territory and regionally-defined winter storm impact models.Results from objective 1 revealed that winter storm impact models can be successfully developed and validated using a number of environmental and dynamic covariates. Results also showed that reducing the initial covariate set resulted in similar predictive accuracy as the full model, thus allowing for faster model runtimes and easier maintenance. In addition, it was found that meteorological variables are the most important for predicting winter storm-related damage.Results from objective 2 showed that regional models could be developed with reasonable accuracy. Similar to objective 1, reduced versions of the models performed better in many cases, indicating that a large number of covariates could be removed. In addition, meteorological variables such as ice, snow, and wind parameters were found to be the most influential for predicting winter storm related damage. Lastly, it was found that for most models, medium-scale and large-scale events (i.e. orders > 50) were predicted better than small-scale events.Lastly, objective 3 compared the service territory and regional models using a case study approach. Results showed that for most operating companies, the service territory model performed better. This was likely due to the large sample size when using data from all operating companies. In addition, the varying sizes of each operating company further limited sample size, thus resulting in lower accuracy for regional models. Despite the service territory model performing better in most cases, this comparison did shed light on the importance of regional models, and their future refinement to increase model performance. Expanding the study period could improve regional models. Overall, results showed that the winter storm impact models developed performed well and that the regional models performed better in many circumstances. In addition, these results showed that a large number of covariates can be removed, thus lowering computing time and model maintenance. The results from this research can provide FirstEnergy with a decision-support tool that can be implemented prior to a winter storm event to inform decision makers within the company, thus lowering costs and reducing restoration times.