Enhancing Weather-Related Power Outage Prediction by Event Severity Classification

Feifei Yang,Peter Watson,Emmanouil N Anagnostou,Marika Koukoula

doi:10.1109/access.2020.2983159

Abstract

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.

Highlights

Threats to the security of power systems are classified into four categories: natural, accidental, malicious, and emerging [1]
We evaluated the outage prediction model performance according to several error metrics, including AE, APE, mean absolute percentage error (MAPE), centered root mean square error (CRMSE), R2(coefficient of determination R-squared), and NSE (Nash-Sutcliffe efficiency), which are described in the appendix
This paper describes a technique for improving power distribution grid outage prediction by incorporating storm severity classification of weather-related outage events in an outage prediction models (OPMs)

Summary

INTRODUCTION

Threats to the security of power systems are classified into four categories: natural, accidental, malicious, and emerging [1]. F. Yang et al.: Enhancing Weather-Related Power Outage Prediction by Event Severity Classification predicting power outages by dividing the training dataset into subsets of events representative of the tested event’s severity could significantly reduce the bias. In Yang et al.’s previous research actual outages information of the severe event was used to categorize that event and selected historical events in that category were used to train the OPM The limitation of this approach is that the method cannot apply for forecasting since the actual outages of an upcoming event are unknown. We demonstrated through a storm-based holdout test that the event severity classification improved outage prediction model (OPM) performance by dividing the training dataset into subsets and selecting optimization coefficients for the predicted events. We demonstrated that the severity classification of forecasted events would significantly reduce the bias of the real-time OPM-based forecasting system

STUDY AREA AND DATA

OUTAGE PREDICTION MODEL

CONDITIONED OPM

RESULTS AND DISCUSSIONS

CONCLUSION

CLASSIFICATION METRICS