SavedAdd your journal

Journals / Mühendislik Bilimleri ve Tasarım Dergisi / 2026 / Cilt: 14 Sayı: 2

A PREDICTIVE MAINTENANCE FRAMEWORK FOR DAY-AHEAD TRANSFORMER OUTAGE RISK ESTIMATION USING HETEROGENEOUS DATA AND TEMPORAL MODELS

HETEROJEN VERİ KAYNAKLARI VE ZAMANSAL MODELLER İLE GÜN ÖNCESİ TRANSFORMATÖR KESİNTİ RİSKİNİN TAHMİNİ İÇİN BİR ÖNGÖRÜCÜ BAKIM ÇERÇEVESİ

Tugay Eren Güzelyol(MANİSA CELÂL BAYAR ÜNİVERSİTESİ)0009-0008-2216-3181
Yavuz Ateş(YILDIZ TEKNİK ÜNİVERSİTESİ)0000-0002-4168-0861
Pages
318–329
Publication Date
2026-06-30
🏛️Bibliographic metadata and full text access for this article are provided via DergiPark.
DergiPark Record ↗
✨

AI Summary & Key Takeaways

AI Insight
⚡ TL;DR(Too Long; Didn't Read)

The findings suggest that the integration of weather and asset data enhances predictive performance and supports proactive maintenance strategies for improving SAIDI and SAIFI reliability metrics.

📌 3 Key Takeaways

🎯
Objective & Hypothesis:Unplanned outages in power distribution networks significantly affect reliability and increase operational costs.
📊
Key Findings & Data:The dataset comprises 8,671 outage records (2021–2024) across 34 transformers, enriched with meteorological data and asset-specific metadata.
💡
Conclusion & Impact:The findings suggest that the integration of weather and asset data enhances predictive performance and supports proactive maintenance strategies for improving SAIDI and SAIFI reliability metrics.
🔬Design:Modeling & Simulation
📍Scope:2021–2024 Verileri
⚙️Variables:Dağıtım Şebekeleri · Kestirimci Bakım · Makine Öğrenmesi

Abstract

Unplanned outages in power distribution networks significantly affect reliability and increase operational costs. This study introduces a predictive maintenance framework for estimating transformer outage risks one day in advance, based on real-world data from İzmir and Manisa. The dataset comprises 8,671 outage records (2021–2024) across 34 transformers, enriched with meteorological data and asset-specific metadata. The proposed pipeline includes outlier detection, temporal feature engineering, and class imbalance handling (SMOTE, undersampling, and class weighting) under Min–Max and Robust scaling. Various machine learning (Random Forest, XGBoost, LightGBM, CatBoost) and deep learning models (MLP, LSTM, GRU, Transformer) were trained with hyperparameter tuning via Optuna. Performance was evaluated using time-series cross-validation and an out-of-sample test set for Q4 of 2024, primarily with the G-Mean metric. Results reveal that gradient boosting models provide balanced accuracy, while sequence-based models are more effective under seasonal variations. The findings suggest that the integration of weather and asset data enhances predictive performance and supports proactive maintenance strategies for improving SAIDI and SAIFI reliability metrics.

Özet

Elektrik dağıtım şebekelerinde meydana gelen plansız kesintiler, şebeke güvenilirliğini azaltmakta ve işletme maliyetlerini artırmaktadır. Bu çalışmada, İzmir ve Manisa illerinden elde edilen gerçek saha verileri kullanılarak, transformatörler için bir gün önceden arıza riski tahmini yapan kestirimci bakım odaklı bir tahminleme modeli önerilmiştir. Çalışmada, 34 transformatöre ait 2021–2024 yılları arasındaki 8.671 kesinti kaydı, bölgesel meteorolojik veriler ve varlık odaklı teknik bilgilerle birleştirilmiştir. Önerilen sistem; aykırı değer tespiti, zaman serisi tabanlı özellik mühendisliği ve dengesiz sınıf problemleri için SMOTE, rastgele örnekleme ve sınıf ağırlıklandırma gibi yöntemleri içermektedir. Ayrıca, Min–Max ve Robust ölçekleme yaklaşımları uygulanmıştır. Makine öğrenmesi algoritmaları (Rastgele Orman, XGBoost, LightGBM, CatBoost) ile derin öğrenme modelleri (MLP, LSTM, GRU, Transformer), Optuna ile hiper parametre ayarı yapılarak eğitilmiştir. Zaman serisi çapraz doğrulama ve 2024 yılının son çeyreği için ayrılmış test seti ile değerlendirilen modelin ana başarı ölçütü G-Mean olmuştur. Sonuçlar, gradyan artırmalı modellerin genel doğruluk açısından üstün olduğunu, sıralı modellerin ise mevsimsellik içeren durumlarda daha başarılı olduğunu göstermektedir. Model, SAIDI ve SAIFI gibi güvenilirlik göstergelerinin iyileştirilmesi için etkili bakım planlaması imkânı sunmaktadır.

🔖

Cite This Article

Güzelyol, T. E., Ateş, Y. (2026). A PREDICTIVE MAINTENANCE FRAMEWORK FOR DAY-AHEAD TRANSFORMER OUTAGE RISK ESTIMATION USING HETEROGENEOUS DATA AND TEMPORAL MODELS. Mühendislik Bilimleri ve Tasarım Dergisi, Cilt: 14 Sayı: 2, 318–329. https://doi.org/10.21923/jesd.1862245

📄 Full Text PDF Viewer

📄

PDF görüntüleyici hazırlanıyor...

References (29)

  1. Akcanca, M. A., & Taşkın, S. (2013). Akıllı Şebeke Uygulanabilirliği Açısından Türkiye Elektrik Enerji Sisteminin İncelenmesi. Akıllı Şebekeler ve Türkiye Elektrik Şebekesinin Geleceği Sempozyumu, 131–135. www.emo.org.tr/ekler/a4d6c5a3223642b_ek.pdf
  2. Campbell, R. J. (2013). Weather-related power outages and electric system resiliency. The Transformation of Electrical Power: Key Issues, 103–118.
  3. Cruz, E., & Munasinghe, T. (2023). Exploring Power Outage Prediction Using Weather and Socioeconomic Data in the Southeastern Part of the United States. Proceedings - 2023 IEEE International Conference on Big Data, BigData 2023, 6138–6142. https://doi.org/10.1109/BigData59044.2023.10386519
  4. Doostan, M., & Chowdhury, B. (2020). Predicting lightning-related outages in power distribution systems: A statistical approach. IEEE Access, 8, 84541–84550. https://doi.org/10.1109/ACCESS.2020.2991923
  5. Fatima, K., Shareef, H., Costa, F. B., Bajwa, A. A., & Wong, L. A. (2024). Machine learning for power outage prediction during hurricanes: An extensive review. Engineering Applications of Artificial Intelligence, 133(PB), 108056. https://doi.org/10.1016/j.engappai.2024.108056
  6. Forsberg, S., Thomas, K., & Bergkvist, M. (2023). Power grid vulnerability analysis using complex network theory: A topological study of the Nordic transmission grid. Physica A: Statistical Mechanics and its Applications, 626, 129072. https://doi.org/10.1016/j.physa.2023.129072
  7. Gamel, S. A., Ghoneim, S. S. M., & Sultan, Y. A. (2024). Improving the accuracy of diagnostic predictions for power transformers by employing a hybrid approach combining SMOTE and DNN. Computers and Electrical Engineering, 117(April), 109232. https://doi.org/10.1016/j.compeleceng.2024.109232
  8. Jian, L., & Tianyuan, T. (2015). LS-SVM based substation circuit breakers maintenance scheduling optimization. International Journal of Electrical Power and Energy Systems, 64, 1251–1258. https://doi.org/10.1016/j.ijepes.2014.09.013
  9. Jin-Qiang, C. (2019). Fault Prediction of a Transformer Bushing Based on Entropy Weight TOPSIS and Gray Theory. Computing in Science and Engineering, 21(6), 55–62. https://doi.org/10.1109/MCSE.2018.2882357
  10. Norouzi, H. (2025). Flexible assessment of line polymer insulators based on condition indexes. Electric Power Systems Research, 238(October 2024), 111145. https://doi.org/10.1016/j.epsr.2024.111145
  11. Papic, M., Agarwal, S., Bian, J., Billinton, R., Dent, C., Dobson, I., Jirutitijaroen, P., Li, W., Menten, T., Mitra, J., Schneider, A., Singh, C., & Vadlamudi, V. V. (2014). Effects of dependent and common mode outages on the reliability of bulk electric system - Part I: Basic concepts. IEEE Power and Energy Society General Meeting, 2014-Octob(October). https://doi.org/10.1109/PESGM.2014.6938967
  12. Qian, L., Qingzhao, T., Qinghui, Z., & Tianqi, W. (2020). Research on power transformer fault prediction model based on LSTM neural network. Proceedings - 2020 International Conference on Robots and Intelligent Systems, ICRIS 2020, 702–705. https://doi.org/10.1109/ICRIS52159.2020.00175
  13. Ren, H., Hou, Z. J., Ke, X., Huang, Q., & Makatov, Y. (2021). Analysis of Weather and Climate Extremes Impact on Power System Outage. IEEE Power and Energy Society General Meeting, 2021-July, 1–5. https://doi.org/10.1109/PESGM46819.2021.9637938
  14. Rexhepi, V. (2017). An Analysis of Power Transformer Outages and Reliability Monitoring. Energy Procedia, 141, 418–422. https://doi.org/10.1016/j.egypro.2017.11.053
  15. Taylor, W. O., Cerrai, D., Wanik, D., Koukoula, M., & Anagnostou, E. N. (2023a). Community power outage prediction modeling for the Eastern United States. Energy Reports, 10(August), 4148–4169. https://doi.org/10.1016/j.egyr.2023.10.073
  16. Taylor, W. O., Cerrai, D., Wanik, D., Koukoula, M., & Anagnostou, E. N. (2023b). Community power outage prediction modeling for the Eastern United States. Energy Reports, 10(October), 4148–4169. https://doi.org/10.1016/j.egyr.2023.10.073
  17. The International Energy Agency (IEA). (2021). No Title. https://www.iea.org/data-and-statistics/charts/world-electricity-final-consumption-by-sector-1974-2019.
  18. Thomas, J. B., & K.V., S. (2023). Neural architecture search algorithm to optimize deep Transformer model for fault detection in electrical power distribution systems. Engineering Applications of Artificial Intelligence, 120(July 2022), 105890. https://doi.org/10.1016/j.engappai.2023.105890
  19. Toubeau, J. F., Pardoen, L., Hubert, L., Marenne, N., Sprooten, J., De Grève, Z., & Vallée, F. (2022). Machine learning-assisted outage planning for maintenance activities in power systems with renewables. Energy, 238. https://doi.org/10.1016/j.energy.2021.121993
  20. Tuttokmaği, Ö., & Kaygusuz, A. (2019). Araştırma Makalesi / Research Article Büyük Ölçekli Elektrik Kesintilerinin İncelenmesi Investigation of Large Scale Power Blackouts. 8(2), 664–676.
  21. Ucar, A., Karakose, M., & Kırımça, N. (2024). applied sciences Artificial Intelligence for Predictive Maintenance Applications
  22. Visakh, A., & Selvan, M. P. (2023). Analysis and mitigation of the impact of electric vehicle charging on service disruption of distribution transformers. Sustainable Energy, Grids and Networks, 35, 101096. https://doi.org/10.1016/j.segan.2023.101096
  23. Wang, S., Sun, H., Kim, K. J., Guo, J., & Nikovski, D. (2021). Solving Bernoulli Bandit Problems for Weather-relative Overhead Distribution Line Failures Forecasting. IEEE Power and Energy Society General Meeting, 2021-July, 0–4. https://doi.org/10.1109/PESGM46819.2021.9638153
  24. Xu, F., Zhou, L., Zhou, Y., Ai, X., & Liu, Q. (2025). Power outage data management platform based on big data analysis technology. Procedia Computer Science, 261, 976–982. https://doi.org/10.1016/j.procs.2025.04.489
  25. Xu, Z., Liu, J., Fan, W., Wang, Y., Luan, L., & Zhou, K. (2022). An Power Outage Prediction Method Based on XGBoost with Improved Objective Function. Proceedings - 2022 37th Youth Academic Annual Conference of Chinese Association of Automation, YAC 2022, 1560–1565. https://doi.org/10.1109/YAC57282.2022.10023921
  26. Yuyama, A., Kajitani, Y., & Shoji, G. (2018). Simulation of operational reliability of thermal power plants during a power crisis: Are we underestimating power shortage risk? Applied Energy, 231(February), 901–913. https://doi.org/10.1016/j.apenergy.2018.09.089
  27. Zaldivar, D. A., Sanchez, A. M., & Romero, A. A. (2023). A comprehensive methodology for the optimization of condition-based maintenance in power transformer fleets. Electric Power Systems Research, 220(March). https://doi.org/10.1016/j.epsr.2023.109374
  28. Zhao, T., Satoshi, E., Yue, M., Jensen, M., Marschilok, A., Nugent, B., Cerruti, B., & Spanos, C. (2023). On Learning-Based Model for Dynamic Granular Prediction of Power Outages Under Extreme Events. 2023 IEEE PES Innovative Smart Grid Technologies Latin America, ISGT-LA 2023, 445–449. https://doi.org/10.1109/ISGT-LA56058.2023.10328277
  29. Zhou, Q., Zhang, J., Yang, Z., & Yang, Y. (2014). Distribution network outage pre-warning analytics under extreme weather conditions. China International Conference on Electricity Distribution, CICED, 2014-Decem(Ciced), 1190–1194. https://doi.org/10.1109/CICED.2014.6991895