Reliability-Oriented Spatiotemporal Machine Learning for High-Impact Power Outage Event Prediction

Authors

• Marwa Gamal, Electrical Engineering Department, Faculty of Engineering, Suez Canal University, Ismailia 41522, Egypt,Follow

Corresponding Author

Marwa Gamal

Subject Area

Computer and Control Systems Engineering

Article Type

Original Study

Abstract

Power outages have become an increasing concern for modern power systems due to their impact on infrastructure reliability, economic activities, and public safety. The growing frequency of extreme weather events and the rising demand for electricity have made it more difficult to anticipate high-impact outage events. One of the main challenges in this context is the complex interaction between temporal patterns and geographic variations, which traditional methods often fail to capture effectively. This study develops a machine learning framework that combines temporal characteristics with geographic information to improve the prediction of high-impact power outages. Temporal features such as seasonal patterns, daily variations, and weekly behaviour are considered alongside spatial attributes including location-based information. This integration enables the models to better represent variations in outage behaviour across both time and location. Several supervised learning models are implemented and evaluated using a real-world dataset of power outages. The assessment focuses not only on accuracy but also on metrics that better reflect reliability, such as recall, F1-score, and the area under the ROC curve. The results show clear differences in model performance when evaluated beyond accuracy alone, particularly in detecting high-impact outage events. Among the tested models, Random Forest demonstrates the most consistent performance, indicating its ability to capture complex relationships within the data. The findings highlight the importance of combining spatial and temporal information in outage prediction and emphasize the need to rely on evaluation measures that reflect practical system reliability rather than accuracy alone.

Keywords

Power outage, Machine learning, Ensemble Learning, Spatiotemporal data, Geographic data.

Creative Commons License

This work is licensed under a Creative Commons Attribution 4.0 License.

Recommended Citation

Gamal, Marwa (2026) "Reliability-Oriented Spatiotemporal Machine Learning for High-Impact Power Outage Event Prediction," Mansoura Engineering Journal: Vol. 51 : Iss. 5 , Article 11.
Available at: https://doi.org/10.58491/2735-4202.3489