Journals / Balkan Journal of Electrical and Computer Engineering / 2026 / Volume: 14

Forecasting High Electricity Demand Using Explainable Machine Learning

Yüksek Elektrik Talebinin Açıklanabilir Makine Öğrenmesiyle Öngörülmesi

Süleyman Dal(BATMAN ÜNİVERSİTESİ)0000-0002-4564-8076
Pages
—
Publication Date
2026-11-13
🏛️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 indicate that the approach could contribute to monitoring residential demand; however, the burden of false alarms and the limited success in duration prediction must be taken into account from a practical standpoint.

📌 3 Key Takeaways

🎯
Objective & Hypothesis:The objective of this study was to predict the onset of high demand, the highest half-hourly average power, and the total duration of high demand over the next two hours using historical electricity consumption data from residential households.
📊
Key Findings & Data:Half-hourly measurements from 2013 for 200 standard-tariff residential households selected from the Low Carbon London dataset were used.
💡
Conclusion & Impact:The findings indicate that the approach could contribute to monitoring residential demand; however, the burden of false alarms and the limited success in duration prediction must be taken into account from a practical standpoint.
🔬Design:Modeling & Simulation
📍Scope:2013 Verileri
⚙️Variables:Smart grids · residential electricity demand · smart meters

Abstract

The objective of this study was to predict the onset of high demand, the highest half-hourly average power, and the total duration of high demand over the next two hours using historical electricity consumption data from residential households. Half-hourly measurements from 2013 for 200 standard-tariff residential households selected from the Low Carbon London dataset were used. High demand was defined as exceeding the 90th percentile of each household’s average power values during the January–February period. The models were developed using data from 160 households; transfer performance was evaluated on the 40 households not included in model training. The process was divided into a training period (March–June), model selection in July, calibration of prediction intervals in August, and a testing period (September–December). The selected models were retrained using March–July data. Separate models were built for the three targets; Random Forest, histogram-based gradient boosting (HistGB), XGBoost, LightGBM, and CatBoost were compared against a weekly seasonal reference. Model selection was initially based on average precision (AP) for the target and mean absolute error (MAE) for power and duration. For the residential units used in the training, the HistGB model selected during the testing period produced an AP of 0.4183; the CatBoost models produced an MAE of 0.2704 kW for power and 9.2361 minutes for duration. The contributions of historical consumption and calendar features to the model predictions were examined using the SHAP method. In the early warning evaluation, 78.81% of relevant events in known residential units were detected in advance; 60.93% of early warnings were identified as false alarms. The improvement in duration prediction relative to a reference that yields a duration of zero in every window remained limited. The findings indicate that the approach could contribute to monitoring residential demand; however, the burden of false alarms and the limited success in duration prediction must be taken into account from a practical standpoint.

Özet

Bu çalışmada, konutların geçmiş elektrik tüketiminden yararlanılarak sonraki iki saatte yüksek talep başlangıcının, en yüksek yarım saatlik ortalama gücün ve yüksek talepte geçirilen toplam sürenin öngörülmesi amaçlanmıştır. Low Carbon London veri setinden seçilen 200 standart tarifeli konutun 2013 yılına ait yarım saatlik ölçümleri kullanılmıştır. Yüksek talep, her konutun Ocak–Şubat dönemindeki ortalama güç değerlerinin %90 yüzdeliğinin aşılması olarak tanımlanmıştır. Modeller 160 konutun verileriyle geliştirilmiş; model eğitimine katılmayan 40 konutta aktarım performansı değerlendirilmiştir. Mart–Haziran eğitim, Temmuz model seçimi, Ağustos tahmin aralıklarının kalibrasyonu ve Eylül–Aralık test dönemi olarak ayrılmıştır. Seçilen modeller Mart–Temmuz verileriyle yeniden eğitilmiştir. Üç hedef için ayrı modeller kurulmuş; Random Forest, histogram tabanlı gradyan artırma (HistGB), XGBoost, LightGBM ve CatBoost haftalık mevsimsel referansla karşılaştırılmıştır. Model seçimi başlangıç için ortalama kesinliğe (AP), güç ve süre için ortalama mutlak hataya (MAE) dayanmıştır. Eğitimde kullanılan konutların test döneminde seçilen HistGB modeli başlangıç için 0,4183 AP; CatBoost modelleri güç için 0,2704 kW ve süre için 9,2361 dakika MAE üretmiştir. SHAP yöntemiyle geçmiş tüketim ve takvim özelliklerinin model tahminlerine katkıları incelenmiştir. Erken uyarı değerlendirmesinde bilinen konutlardaki uygun olayların %78,81’i önceden yakalanmış; erken uyarıların %60,93’ü yanlış uyarı olarak belirlenmiştir. Süre tahmininin her pencerede sıfır süre veren referansa göre iyileşmesi sınırlı kalmıştır. Bulgular, yaklaşımın konut talebinin izlenmesine katkı sağlayabileceğini, ancak yanlış uyarı yükünün ve süre tahminindeki sınırlı başarının uygulama açısından dikkate alınması gerektiğini göstermektedir.

🔖

Cite This Article

Dal, S. (2026). Forecasting High Electricity Demand Using Explainable Machine Learning. Balkan Journal of Electrical and Computer Engineering, Volume: 14. https://doi.org/10.17694/bajece.2060791

📄 Full Text PDF Viewer

📄

PDF görüntüleyici hazırlanıyor...

References (17)

  1. Richardson, I., Thomson, M., Infield, D., & Clifford, C. (2010). Domestic electricity use: A high-resolution energy demand model. Energy and Buildings, 42(10), 1878–1887.
  2. Satre-Meloy, A., Diakonova, M., & Grünewald, P. (2020). Daily life and demand: An analysis of intra-day variations in residential electricity consumption with time-use data. Energy Efficiency, 13(3), 433–458.
  3. Azarova, V., Engel, D., Ferner, C., Kollmann, A., & Reichl, J. (2018). Exploring the impact of network tariffs on household electricity expenditures using load profiles and socio-economic characteristics. Nature Energy, 3(4), 317–325.
  4. He, X., & Reiner, D. (2016). Electricity demand and basic needs: Empirical evidence from China's households. Energy Policy, 90, 212–221.
  5. Shi, G., Zheng, X., & Song, F. (2012). Estimating elasticity for residential electricity demand in China. The Scientific World Journal, 2012(1), 395629.
  6. UK Power Networks. (n.d.). SmartMeter energy consumption data in London households [Data set]. London Datastore. https://data.london.gov.uk/dataset/smartmeter-energy-consumption-data-in-london-households-vqm0d
  7. Hyndman, R. J., & Fan, Y. (1996). Sample quantiles in statistical packages. The American Statistician, 50(4), 361–365.
  8. García, S., Luengo, J., & Herrera, F. (2015). Data preprocessing in data mining. Springer.
  9. Dal, S., & Sezgin, N. (2025). Heart attack classification with a machine learning approach based on the random forest algorithm. Balkan Journal of Electrical and Computer Engineering, 13(2), 140–147.
  10. Hastie, T., Tibshirani, R., & Friedman, J. (2009). The elements of statistical learning: Data mining, inference, and prediction (2nd ed.). Springer.
  11. Fouladvand, S., Noshad, M., Goldstein, M. K., Periyakoil, V. J., & Chen, J. H. (2023). Mild cognitive impairment: Data-driven prediction, risk factors, and workup. AMIA Summits on Translational Science Proceedings, 2023, 167.
  12. Park, J., & Hwang, E. (2021). A two-stage multistep-ahead electricity load forecasting scheme based on LightGBM and attention-BiLSTM. Sensors, 21(22), 7697.
  13. Dal, S. (2025). Eye state classification from electroencephalography (EEG) signals using the extra trees classifier algorithm. European Journal of Technique (EJT), 15(1), 29–36.
  14. 14] Dal, S., & Sezgın, N. (2024). Optimal parameter extraction of triple-diode photovoltaic model using frilled lizard optimization. In 2024 Global Energy Conference (GEC) (pp. 1–5). IEEE.
  15. Lundberg, S. M., Erion, G., Chen, H., DeGrave, A., Prutkin, J. M., Nair, B., Katz, R., Himmelfarb, J., Bansal, N., & Lee, S.-I. (2020). From local explanations to global understanding with explainable AI for trees. Nature Machine Intelligence, 2(1), 56–67.
  16. Ke, G., Meng, Q., Finley, T., Wang, T., Chen, W., Ma, W., Ye, Q., & Liu, T.-Y. (2017). LightGBM: A highly efficient gradient boosting decision tree. Advances in Neural Information Processing Systems, 30.
  17. Prokhorenkova, L., Gusev, G., Vorobev, A., Dorogush, A. V., & Gulin, A. (2018). CatBoost: Unbiased boosting with categorical features. Advances in Neural Information Processing Systems, 31.