Gust Rectification and Stacking Based Methodology for Power Outage Prediction under Hurricanes

Shiwen Yu,Lina He

DOI: https://doi.org/10.1109/pesgm46819.2021.9638182

2021-01-01

Abstract:The threat of hurricanes to the security and resiliency of power systems is increasing with climate change and global warming. Due to the complex relationship between the hurricane disaster and power outages, it is difficult for power system operators to predict outages accurately and allocate emergency supplies in advance. Therefore, establishing an efficient and effective approach to forecast power outage distribution is critical for hurricane mitigation and loss reduction. This paper proposes a machine-learning based prediction methodology based on the realistic historical hurricane outage data. The proposed methodology includes two parts: (1) rectifying gust via the power law, and (2) configuring stacked model based on machine learning algorithms and metrics optimization. In particular, the proposed gust rectification algorithm can improve the interpolation precision effectively. By implementing this configuration, the stacked model has a much better performance than the single model in the stacking processing. This model is trained by a comprehensive data set that consists of the information of hurricanes, power systems, and geography. In the case study, the realistic data of two historical hurricanes is applied to verify the developed model. The results demonstrate that the proposed methodology can improve the performance of power outage prediction effectively.