Makine Öğrenmesi Algoritmaları ile Elektrik Dağıtım Şebekeleri Arıza Tahmini

Yıl 2025, Cilt: 15 Sayı: 1, 73 - 98, 15.03.2025

Ali Geyikoğlu , Mete Yağanoğlu

https://doi.org/10.31466/kfbd.1482179

Öz

Elektrik dağıtım şebekelerinde arıza; kaliteli ve sürekli enerji akışını engelleyici faktörler olarak tanımlanmaktadır. Arızanın meydana gelmesi sonrasında Elektrik Dağıtım Şirketleri, bakım-onarım ve yatırım çalışmaları ile düzeltici faaliyetler gerçekleştirmektedir. Meydana gelen arızalar ve sonrası düzeltici faaliyetler ile teknik kalite parametreleri sistemlerce oluşturulmaktadır. Ancak ortaya çıkan teknik veriler, herhangi bir tahminleme altyapısında kullanılmamakta, düzeltici faaliyetler genel olarak yorum ve taleplere istinaden gerçekleştirilmektedir. Bu çalışmada, sezgisel yaklaşımların önüne geçmek amacıyla, elektrik dağıtım şirketi operatörlerinin saha faaliyetleri sonrası sistemler tarafından örneklenerek kayıt altına alınan Aras EDAŞ’a ait Kesinti Süreleri ve Sıklığı verileri ile ilgili dönemlere ait Aras EDAŞ işletme sorumluluk sahasındaki 7 ile esas meteorolojik veriler kullanılmıştır. Veri seti içerisinde yer alan öznitelikler ve sınıflar üzerinde veri ön işleme, öznitelik seçimi, öznitelik çıkarımı gerçekleştirilmiştir. Regresyon işlemleri ile tahminleme gerçekleştirilecek hale gelen veri setleri %80’i eğitim ve %20’si test verisi olacak şekilde; Hafif Gradyan Artırma Makinesi (LGBM), Aşırı Gradyan Artırma (XGB), Destek Vektör, Rastgele Orman, Kategorik Artırma, k-En Yakın Komşu, Karar Ağacı, Lineer olmak üzere 8 farklı regresyon modeline tabi tutulmuştur. Veri seti üzerinde yer alan iki farklı bağımlı değişkene ait çok sınıflı değerler ayrı ayrı sınıf modeline dahil edilmiş olup toplamda 8 farklı model için 16 adet regresyon çalışması gerçekleştirilmiştir. En iyi model yapısına ulaşabilmek amacıyla hiperparametre optimizasyonu uygulanmıştır. Birincil çok sınıflı regresyon tahmini için en iyi model doğruluğu LGBM Regressor ile %93,305 olarak elde edilirken, ikincil çok sınıflı tahmin için en iyi model doğruluğu XGB Regressor ile %95,812 olarak elde edilmiştir.

Anahtar Kelimeler

Elektrik dağıtım şebeke arızası, Makine öğrenmesi, Regresyon, Tahmin

Electricity Distribution Networks Fault Prediction with Machine Learning Algorithms

Öz

Malfunction in electrical distribution networks; They are defined as factors that prevent quality and continuous energy flow. After the failure occurs, Electricity Distribution Companies carry out corrective actions through maintenance-repair and investment works. Failures that occur and subsequent corrective actions and technical quality parameters are created by the systems. However, the resulting technical data is not used in any forecasting infrastructure, and corrective actions are generally carried out based on comments and requests. In this study, in order to avoid intuitive approaches, the Interruption Duration and Frequency data of Aras EDAŞ, which were sampled and recorded by the systems after the field activities of the electricity distribution company operators, and the main meteorological data of 7 provinces in the Aras EDAŞ operational responsibility area for the relevant periods were used. Data preprocessing, feature selection, and feature extraction were carried out on the attributes and classes in the data set. The data sets that will be used for estimation with regression operations were subjected to 8 different regression models, including Light Gradient Boosting Machine (LGBM), Extreme Gradient Boosting (XGB), Support Vector, Random Forest, Categorical Boosting, k-Nearest Neighbor, Decision Tree, and Linear, with 80% of the data being training and 20% being test data. Multi-class values of two different dependent variables on the data set were included separately in the class model, and a total of 16 regression studies were carried out for 8 different models. Hyperparameter optimization was applied to achieve the best model structure. While the best model accuracy for primary multi-class regression prediction was obtained as 93.305% with the LGBM Regressor, the best model accuracy for secondary multi-class prediction was obtained as 95.812% with the XGB Regressor.

Anahtar Kelimeler

Electricity distribution network failure, Machine learning, Regression, Prediction

Kaynakça

• Abdel-Nasser, M., Mahmoud, K., & Kashef, H. (2018). A novel smart grid state estimation method based on neural networks. IJIMAI, 5(1), 92-100.
• Beskopylny, A. N., Stel’makh, S. A., Shcherban’, E. M., Mailyan, L. R., Meskhi, B., Razveeva, I., ... & Beskopylny, N. (2022). Concrete strength prediction using machine learning methods CatBoost, k-Nearest Neighbors, Support Vector Regression. Applied Sciences, 12(21), 10864.
• Chen, T., & Guestrin, C. (2016, August). Xgboost: A scalable tree boosting system. In Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining (pp. 785-794).
• Dashtdar, M., Dashti, R., & Shaker, H. R. (2018, May). Distribution network fault section identification and fault location using artificial neural network. In 2018 5th international conference on electrical and electronic engineering (ICEEE) (pp. 273-278). IEEE.
• De Santis, E., Mascioli, F. M. F., Sadeghian, A., & Rizzi, A. (2016). A dissimilarity learning approach by evolutionary computation for faults recognition in smart grids. In Computational Intelligence: International Joint Conference, IJCCI 2014 Rome, Italy, October 22-24, 2014 Revised Selected Papers (pp. 113-130). Springer International Publishing.
• Drucker, H., Burges, C. J., Kaufman, L., Smola, A., & Vapnik, V. (1996). Support vector regression machines. Advances in neural information processing systems, 9.
• Erdem, E., & Karamustafaoğlu, M., Elektrik Dağıtım Sektör Raporu. (2021, 31 Aralık) https://www.erdem-erdem.av.tr/bilgi-bankasi/elektrik-dagitim-sektor-raporu
• Ferreira, A. B., Leite, J. B., & Salvadeo, D. H. (2025). Power substation load forecasting using interpretable transformer-based temporal fusion neural networks. Electric Power Systems Research, 238, 111169.
• Graw, J. H., Wood, W. T., & Phrampus, B. J. (2021). Predicting global marine sediment density using the random forest regressor machine learning algorithm. Journal of Geophysical Research: Solid Earth, 126(1), e2020JB020135.
• Guo, J., Yun, S., Meng, Y., He, N., Ye, D., Zhao, Z., ... & Yang, L. (2023). Prediction of heating and cooling loads based on light gradient boosting machine algorithms. Building and Environment, 236, 110252.
• Hassani, H., Razavi–Far, R., & Saif, M. (2019, October). Locating faults in smart grids using neuro–fuzzy networks. In 2019 IEEE International Conference on Systems, Man and Cybernetics (SMC) (pp. 3281-3286). IEEE.
• Jamali, S., Bahmanyar, A., & Ranjbar, S. (2020). Hybrid classifier for fault location in active distribution networks. Protection and Control of Modern Power Systems, 5, 1-9.
• James, G., Witten, D., Hastie, T., Tibshirani, R., & Taylor, J. (2023). Linear regression. In An Introduction to Statistical Learning: With Applications in Python (pp. 69-134). Cham: Springer International Publishing.
• Kankanala, P., Das, S., & Pahwa, A. (2013). AdaBoost $^{+} $: An Ensemble Learning Approach for Estimating Weather-Related Outages in Distribution Systems. IEEE Transactions on Power Systems, 29(1), 359-367.
• Kurup, A. R., Martínez–Ramón, M., Summers, A., Bidram, A., & Reno, M. J. (2021, October). Deep learning-based circuit topology estimation and fault classification in distribution systems. In 2021 IEEE PES Innovative Smart Grid Technologies Europe (ISGT Europe) (pp. 01-05). IEEE.
• Mai, T. T., Nguyen, P. H., Haque, N. A., & Pemen, G. A. (2022). Exploring regression models to enable monitoring capability of local energy communities for self‐management in low‐voltage distribution networks. IET Smart Grid, 5(1), 25-41.
• Majidi, M., Etezadi-Amoli, M., & Fadali, M. S. (2014). A novel method for single and simultaneous fault location in distribution networks. IEEE Transactions on Power Systems, 30(6), 3368-3376.
• Mestav, K. R., & Tong, L. (2019, October). State estimation in smart distribution systems with deep generative adversary networks. In 2019 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids (SmartGridComm) (pp. 1-6). IEEE.
• Mitchell, T. M., & Mitchell, T. M. (1997). Machine learning (Vol. 1, No. 9). New York: McGraw-hill.
• Montgomery, D. C., Peck, E. A., & Vining, G. G. (2021). Introduction to linear regression analysis. John Wiley & Sons.
• Mori, H., & Yokoyama, H. (2016). A hybrid intelligent method for estimating distribution network reconfigurations. IFAC-PapersOnLine, 49(27), 152-157.
• Mori, H., Aoyama, H., Yamanaka, T., & Urano, S. (2002, October). A fault detection technique with preconditioned ANN in power systems. In IEEE/PES transmission and distribution conference and exhibition (Vol. 2, pp. 758-763). IEEE.
• Niranjan, T., Swetha, D., Charitha, V., & Stephen, A. J. (2019). Predicting Burned Area Of Forest Fires. IRJCS:: International Research Journal of Computer Science, 6, 132-136.
• Nirmal, M. S., Patil, P., & Kumar, J. R. R. (2024). CNN-AdaBoost based Hybrid Model for Electricity Theft Detection in Smart Grid. e-Prime-Advances in Electrical Engineering, Electronics and Energy, 100452.
• Parbat, D., & Chakraborty, M. (2020). A python-based support vector regression model for prediction of COVID19 cases in India. Chaos, Solitons & Fractals, 138, 109942.
• PARLAK, B. O., & YAVAŞOĞLU, H. A. (2023). Comparison of Regression Algorithms to Predict Average Air Temperature. International Journal of Engineering Research and Development, 15(1), 312-322.
• Perles, A., Camilleri, G., & Gleizes, M. P. (2017). Self-adaptive distribution system state estimation. In Multi-Agent Systems and Agreement Technologies: 14th European Conference, EUMAS 2016, and 4th International Conference, AT 2016, Valencia, Spain, December 15-16, 2016, Revised Selected Papers (pp. 202-216). Springer International Publishing.
• Song, Y., Liang, J., Lu, J., & Zhao, X. (2017). An efficient instance selection algorithm for k nearest neighbor regression. Neurocomputing, 251, 26-34.
• T.C. EPDK-Enerji Piyasası Düzenleme Kurumu. (Erişim Tarihi: 2 Mayıs 2023). https://www.epdk.gov.tr
• Uddin, S., Khan, A., Hossain, M. E., & Moni, M. A. (2019). Comparing different supervised machine learning algorithms for disease prediction. BMC medical informatics and decision making, 19(1), 1-16.
• Xinrui, L., Yaoyao, Z., Peng, J., & Tianqi, L. (2017, May). Analysis of ice disaster failure considering the multi angle information modification for distribution network. In 2017 29th Chinese Control And Decision Conference (CCDC) (pp. 6685-6690). IEEE.
• Xu, M., Watanachaturaporn, P., Varshney, P. K., & Arora, M. K. (2005). Decision tree regression for soft classification of remote sensing data. Remote Sensing of Environment, 97(3), 322-336.
• Yellagoud, S. K., & Talluri, P. R. (2019). A comparative evaluation of AI based fault location tools for electric distribution networks. Int. J. Power Energy Syst, 39(4), 177-183.
• Zhou, R., Li, Y., & Lin, X. (2025). A clustered federated learning framework for collaborative fault diagnosis of wind turbines. Applied Energy, 377, 124532.