Akıllı Şebekelerde Elektrik Hırsızlığı Tespiti İçin Gelişmiş Makine Öğrenmesi Yöntemlerinin Uygulanması

Yıl 2025, Cilt: 40 Sayı: 3, 627 - 641, 26.09.2025

Ömer Can Tolun , Kasım Zor , Önder Tutsoy

https://doi.org/10.21605/cukurovaumfd.1772073

Öz

Elektrik hırsızlığı, dağıtım sistemlerinde hem ekonomik kayıplara hem de şebeke güvenilirliğinin azalmasına yol açan en kritik sorunlardan biridir. Bu çalışmada, elektrik hırsızlığı tespiti (EHT) için veri dengesizliği ve model iyileştirme problemlerini aynı anda ele alan gelişmiş bir makine öğrenmesi algoritması sunulmaktadır. Öncelikle, meskenlerden elde edilen gerçek elektrik tüketim verilerine ön işleme adımları uygulandıktan sonra dengesizliği azaltmak için K-ortalamalar tabanlı bir dengeleme yöntemi uygulanmıştır. Boyut indirgeme aşamasında, modelin yalnızca en anlamlı girdilerle eğitilmesi amacıyla karşılıklı bilgi temelli en iyilerin seçilmesi yöntemi tercih edilmiştir. Sınıflandırma aşamasında hafif gradyan artırmalı makine (LightGBM) algoritması kullanılmış ve tahmin performansı Bayes optimizasyon yöntemi ile iyileştirilmiştir. Test verisi üzerinde yapılan değerlendirmelerde, önerilen yöntemin sırasıyla %89,1 doğruluk, %97,2 kesinlik, %80,6 duyarlılık ve %88,1 F1 skoru ile güçlü sonuçlar verdiği görülmüştür.

Anahtar Kelimeler

Bayes optimizasyon , boyut indirgeme , elektrik hırsızlığı tespiti , LightGBM

Application of Advanced Machine Learning Methods for Electricity Theft Detection in Smart Grids

Öz

Electricity theft is one of the most critical issues in distribution systems causing both economic losses and reduced network reliability. In this study, an advanced machine learning framework is proposed for electricity theft detection along with addressing data imbalance and model improvement problems simultaneously. Firstly, after applying preprocessing steps to the actual electricity consumption data obtained from households, a K-means-based balancing method was employed to reduce data imbalance. In the dimension reduction stage, the SelectKBest method based on mutual information was preferred to ensure that the model was trained with only the most significant inputs. In the classification phase, the light gradient boosting machine (LightGBM) algorithm was employed and the prediction performance was enhanced using the Bayesian optimisation method. Evaluations on the test dataset demonstrated that the proposed method achieved strong results with 89.1% accuracy, 97.2% precision, 80.6% recall, and 88.1% F1-score, respectively.

Anahtar Kelimeler

Bayesian optimisation , dimension reduction , electricity theft detection , LightGBM

Kaynakça

• 1. Kebotogetse, O., Samikannu, R. & Yahya, A. (2022). A concealed based approach for secure transmission in advanced metering infrastructure. IEEE Access, 10, 84809-84817.
• 2. Esmael, A.A., Da Silva, H.H., Ji, T. & da Silva Torres, R. (2021). Non-technical loss detection in power grid using information retrieval approaches: A comparative study. IEEE Access, 9, 40635-40648.
• 3. de Souza Savian, F., Siluk, J.C.M., Garlet, T.B., do Nascimento, F.M., Pinheiro, J.R. & Vale, Z. (2021). Non-technical losses: A systematic contemporary article review. Renewable and Sustainable Energy Reviews, 147, 111205.
• 4. Northeast Group, (2021). Electricity theft and non-technical losses: Global markets, solutions, and vendors. https://northeast-group.com/2021/10/20/electricity-theft-non-technical-losses/.
• 5. Tasdoven, H., Fiedler, B.A. & Garayev, V. (2012). Improving electricity efficiency in Turkey by addressing illegal electricity consumption: A governance approach. Energy Policy, 43, 226-234.
• 6. T.C. Enerji Piyasası Düzenleme Kurumu (EPDK), (2024). 2023 yılı elektrik piyasası gelişim raporu. https://www.epdk.gov.tr/Detay/Icerik/4-14475/duyuru. Erişim tarihi: 30 Ağustos 2024.
• 7. Zulu, C.L. & Dzobo, O. (2023). Real-time power theft monitoring and detection system with double connected data capture system. Electrical Engineering, 105(5), 3065-3083.
• 8. Kawoosa, A.I., Prashar, D., Faheem, M., Jha, N. & Khan, A.A. (2023). Using machine learning ensemble method for detection of energy theft in smart meters. IET Generation, Transmission & Distribution, 17(21), 4794-4809.
• 9. Abdulaal, M.J., Ibrahem, M.I., Mahmoud, M.M., Khalid, J., Aljohani, A.J., Milyani, A.H. & Abusorrah, A.M. (2022). Real-time detection of false readings in smart grid AMI using deep and ensemble learning. IEEE Access, 10, 47541-47556.
• 10. Žarković, M. & Dobrić, G. (2024). Artificial intelligence for energy theft detection in distribution networks. Energies, 17(7), 1580.
• 11. Tripathi, A.K., Pandey, A.C. & Sharma, N. (2024). A new electricity theft detection method using hybrid adaptive sampling and pipeline machine learning. Multimedia Tools and Applications, 83(18), 54521-54544.
• 12. Zhuang, W., Jiang, W., Xia, M. & Liu, J. (2024). Dynamic generative residual graph convolutional neural networks for electricity theft detection. IEEE Access, 12, 42737-42750.
• 13. Peng, Y., Yang, Y., Xu, Y., Xue, Y., Song, R., Kang, J. & Zhao, H. (2021). Electricity theft detection in AMI based on clustering and local outlier factor. IEEE Access, 9, 107250-107259.
• 14. Qi, R., Zheng, J., Luo, Z. & Li, Q. (2022). A novel unsupervised data-driven method for electricity theft detection in AMI using observer meters. IEEE Transactions on Instrumentation and Measurement, 71, 1-10.
• 15. El-Toukhy, A.T., Badr, M.M., Mahmoud, M.M., Srivastava, G., Fouda, M.M. & Alsabaan, M. (2023). Electricity theft detection using deep reinforcement learning in smart power grids. IEEE Access, 11, 59558-59574.
• 16. Chen, J., Nanehkaran, Y.A., Chen, W., Liu, Y. & Zhang, D. (2023). Data-driven intelligent method for detection of electricity theft. International Journal of Electrical Power & Energy Systems, 148, 108948.
• 17. Zhang, W. & Dai, Y. (2024). A multiscale electricity theft detection model based on feature engineering. Big Data Research, 36, 100457.
• 18. Althobaiti, A., Rotsos, C. & Marnerides, A.K. (2023). Adaptive energy theft detection in smart grids using self-learning with dual neural network. IEEE Transactions on Industrial Informatics, 20(2), 2776-2786.
• 19. Liao, W., Yang, D., Ge, L., Jia, Y. & Yang, Z. (2025). Electricity theft detection in integrated energy systems considering multi-energy loads. International Journal of Electrical Power & Energy Systems, 164, 110428.
• 20. Sharma, A. & Tiwari, R. (2024). Anomaly detection in smart grid using optimized extreme gradient boosting with SCADA system. Electric Power Systems Research, 235, 110876.
• 21. Zhu, S., Xue, Z. & Li, Y. (2024). Electricity theft detection in smart grids based on omni-scale cnn and autoxgb. IEEE Access, 12, 15477-15492.
• 22. Norouzi, A., Ahmadi, I., Nazari, M., Pourrostami, H. & Hoseini-Kordkheili, H. (2025). A novel framework to detect electricity theft in regions with traditional metering; applying a hybrid machine learning-based approach to low-sampling-rate data. Results in Engineering, 25, 104478.
• 23. Sun, X., Hu, J., Zhang, Z., Cao, D., Huang, Q., Chen, Z. & Hu, W. (2023). Electricity theft detection method based on ensemble learning and prototype learning. Journal of Modern Power Systems and Clean Energy, 12(1), 213-224.
• 24. Nirmal, S., Patil, P. & Shinde, S. (2025). Adversarial measurements for convolutional neural network-based energy theft detection model in smart grid. e-Prime-Advances in Electrical Engineering, Electronics and Energy, 100909.
• 25. Ullah, A., Khan, I.U., Younas, M.Z., Ahmad, M. & Kryvinska, N. (2025). Robust resampling and stacked learning models for electricity theft detection in smart grid. Energy Reports, 13, 770-779.
• 26. Kılınç, E. (2024). Comparison of feature extraction methods in high dimensional time series. Çukurova Üniversitesi Mühendislik Fakültesi Dergisi, 39(4), 991-997.
• 27. CER Smart Metering Project - Electricity Customer Behaviour Trial, commission for energy regulation (cer), 2009-2010, https://www.ucd.ie/issda/data/commissionforenergyregulationcer/.
• 28. Fei, K., Li, Q., Zhu, C., Dong, M. & Li, Y. (2022). Electricity frauds detection in Low-voltage networks with contrastive predictive coding. International Journal of Electrical Power & Energy Systems, 137, 107715.
• 29. Takiddin, A., Ismail, M., Zafar, U. & Serpedin, E. (2022). Deep autoencoder-based anomaly detection of electricity theft cyberattacks in smart grids. IEEE Systems Journal, 16(3), 4106-4117.
• 30. K. Zor, (2019). Research and application of real-time short-term electrical energy consumption forecasting using artificial intelligence based techniques. Ph.D. dissertation, Cukurova University, Adana.
• 31. Leys, C., Ley, C., Klein, O., Bernard, P. & Licata, L. (2013). Detecting outliers: Do not use standard deviation around the mean, use absolute deviation around the median. Journal of Experimental Social Psychology, 49(4), 764-766.
• 32. Lohrer, A., Kazempour, D., Hünemörder, M. & Kröger, P. (2024). CoMadOut—a robust outlier detection algorithm based on CoMAD. Machine Learning, 113, 3553-3580.
• 33. Xu, Z. & Shen, D. (2021). A cluster-based oversampling algorithm combining SMOTE and k-means (KNSMOTE).
• 34. Vergara, J.R. & Estévez, P.A. (2014). A review of feature selection methods based on mutual information. Neural Computing and Applications, 24(1), 175-186.
• 35. Snoek, J., Larochelle, H. & Adams, R.P. (2012). Practical bayesian optimization of machine learning algorithms. Advances in Neural Information Processing Systems, 25, 2960-2968.
• 36. Bentéjac, C., Csörgő, A. & Martínez-Muñoz, G. (2021). A comparative analysis of gradient boosting algorithms. Artificial Intelligence Review, 54, 1937-1967.
• 37. LightGBM Developers, (2025). LightGBM documentation (Parameters & LGBMClassifier).
• 38. Omotehinwa, T.O., Oyewola, D.O. & Dada, E.G. (2023). A light gradient-boosting machine algorithm with tree-structured parzen estimator for breast cancer diagnosis. Healthcare Analytics, 4, 100218.
• 39. Atalay, B.A. & Zor, K. (2025). Xgboost (aşırı gradyan artırımlı karar ağaçları) ile hidroelektrik enerji tahmini. Çukurova Üniversitesi Mühendislik Fakültesi Dergisi, 40(1), 205-218.
• 40. Yaro, A.S., Maly, F., Prazak, P. & Malý, K. (2024). Outlier detection performance of a modified z-score method in time-series rss observation with hybrid scale estimators. IEEE Access, 12, 12785-12796.