Predicting Thunderstorm-Induced Power Outages to Support Utility Restoration

Kabir, Elnaz; Guikema, Seth D.; Quiring, Steven M.

doi:10.1109/tpwrs.2019.2914214

Cited by 46 publications

(37 citation statements)

References 46 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The learned relationship can also be used to draw insights into the influence of different explanatory factors in x on y (e.g., Rivero-Calle et al 2015). It should be noted that supervised learning methods can produce either point estimates or probability distributions as predictions, with approaches such as quantile regression forests (e.g., Kabir et a. 2019) and Bayesian belief networks trained with past data (e.g., Francis et al 2014) being examples of probabilistic methods.…”

Section: Artificial Intelligence Methods and Their Expanding Reachmentioning

confidence: 99%

Artificial Intelligence for Natural Hazards Risk Analysis: Potential, Challenges, and Research Needs

Guikema

2020

Risk Analysis

View full text Add to dashboard Cite

Artificial intelligence (AI) methods have seen increasingly widespread use in everything from consumer products and driverless cars to fraud detection and weather forecasting. The use of AI has transformed many of these application domains. There are ongoing efforts at leveraging AI for disaster risk analysis. This paper takes a critical look at the use of AI for disaster risk analysis. What is the potential? How is the use of AI in this field different from its use in non-disaster fields? What challenges need to be overcome for this potential to be realized? And what are the potential pitfalls of an AI-based approach for disaster risk analysis that we as a society must be cautious of?

show abstract

Section: Artificial Intelligence Methods and Their Expanding Reachmentioning

confidence: 99%

Artificial Intelligence for Natural Hazards Risk Analysis: Potential, Challenges, and Research Needs

Guikema

2020

Risk Analysis

View full text Add to dashboard Cite

show abstract

“…additive regression trees (BART) to predict outages from storm events, while Eskandarpour and Khodaei [24] explored artificial neural network logistic regression to predict outages from an imminent hurricane. Kabir et al [25] utilized RF, boosting tree (BT), support vector machine (SVM), and quantile regression forest (QRF) for predicting thunderstorm-induced power outages. These multiple ML approaches have pushed forward the ability to predict power outages for electric distribution networks.…”

Section: Introductionmentioning

confidence: 99%

“…Although outage prediction modeling has thus achieved improvements through ML methods, current models [19]- [25] often use the same input training datasets of historical events for predicting events at differing levels of severity. Some previous studies have taken the approach of using diversity sampling with balanced class distribution in their training datasets [26]- [29], few investigates predicting power outages by dividing the training dataset into subsets of events representative of the tested event's severity.…”

Section: Introductionmentioning

confidence: 99%

Enhancing Weather-Related Power Outage Prediction by Event Severity Classification

et al. 2020

View full text Add to dashboard Cite

The accuracy of machine learning-based power outage prediction models (OPMs) is sensitive to how well event severity is represented in their training datasets. Unbalanced or overly dispersed event severity can result in random errors in outage predictions and underestimation in severe events or overestimation in weak ones. To improve accuracy in the prediction of storm-caused power outages, we introduce a novel method called ''Conditioned OPM'' that divides an OPM training dataset into subsets of events representative of the predicted event's severity by calculating the quantile weight distance (QWD) between severe weather-related events in the dataset and the predicted event. Based on 102 storm events (including two hurricanes, Irene and Sandy), that have occurred since 2005 over Eversource Energy's Connecticut service territory, we quantified the weather differences among predicted events, which we classified into three groups of severity: low, moderate, and high. The Conditioned OPM creates a subset of the historical events based on their classified severity group and uses that subset as the training dataset to predict the power outages. The study shows that the accuracy of event severity classification was 0.76, and the mean absolute percentage error (MAPE) decreased by about 30%; this method was also tested on forecast events and exhibited a low (20%) MAPE.

show abstract

“…The authors in [11], [12], and [13] present comprehensive studies on predicting hurricane and storm-related outages by addressing important issues such as the presence of a high imbalance in the response variable, engineering informative predicting variables, and building multi-stage models. It is worth mentioning that the focus of the aforementioned studies is not necessarily on lightning-related outages but rather on storm-related that could include various causes of outages (and their combination) such as wind, vegetation, and lightning, to name a few.…”

Section: Introductionmentioning

confidence: 99%

Predicting Lightning-Related Outages in Power Distribution Systems: A Statistical Approach

Doostan

Chowdhury

2020

IEEE Access

View full text Add to dashboard Cite

This paper presents a novel data-driven approach for predicting lightning-related outages that occur in power distribution systems on a daily basis. In order to develop an approach that is able to successfully fulfill this objective, there are two main challenges that ought to be addressed. The first challenge is to define the extent of the target area. An unsupervised machine learning approach is proposed to overcome this difficulty. The second challenge is to adequately identify characteristics of lightning-related outages and to explore the relationship between these outages and weather-related variables (thunderstorm events). In this paper, these outages are clustered into a few manageable groups. Then, a probabilistic model is presented to estimate the likelihood of each group of outages. Finally, a machine learning classification algorithm that can handle the imbalanced problem is developed to predict what group will the outage belong to on a specific day in a specific area of the system under study. Actual outage data, obtained from a major utility in the U.S., in addition to radar weather forecast data are utilized to build the proposed approach. Also, three case studies are provided to show several issues associated with predicting lightning-related outages, and to demonstrate how the proposed approach can address those problems adequately.

show abstract

Predicting Thunderstorm-Induced Power Outages to Support Utility Restoration

Cited by 46 publications

References 46 publications

Artificial Intelligence for Natural Hazards Risk Analysis: Potential, Challenges, and Research Needs

Artificial Intelligence for Natural Hazards Risk Analysis: Potential, Challenges, and Research Needs

Enhancing Weather-Related Power Outage Prediction by Event Severity Classification

Predicting Lightning-Related Outages in Power Distribution Systems: A Statistical Approach

Contact Info

Product

Resources

About