alternativeАльтернативная энергетика и экология (ISJAEE)Alternative Energy and Ecology (ISJAEE)1608-8298Международный издательский дом научной периодики "Спейс10.15518/isjaee.2026.01.176-194alternative-2772Research ArticleXI. ИННОВАЦИОННЫЕ РЕШЕНИЯ, ТЕХНОЛОГИИ, УСТРОЙСТВА И ИХ ВНЕДРЕНИЕ. 27. Информационные технологииXI. INNOVATION SOLUTIONS, TECHNOLOGIES, FACILITIES AND THEIR INNOVATION. 27. Information technologies (IT)Гибридная модель EMD-двухветвевой MLP с многоцелевой оптимизацацией NSGA-II для прогнозирования нагрузки на сутки вперёдEMD-Dual-Branch MLP with NSGA-II Multi-Objective Optimization for Day-Ahead Load ForecastingЧэньСяоюйаChenXiaoyu

Чэнь Сяоюй, аспирант 

620062, Екатеринбург, ул. Мира, д. 19

Chen Xiaoyu, graduate student

620062, Yekaterinburg, Mira st., 19

schen@urfu.ruВелькинВ. И.VelkinV. I.

Велькин Владимир Иванович, профессор Уральского Федерального Университета, кафедра атомных станций и возобновляемых источников энергии

620062, Екатеринбург, ул. Мира, д. 19

Velkin Vladimir Ivanovic, Professor of the Ural Federal University, Department of nuclear power plants and renewable energy sources

620062, Yekaterinburg, Mira st., 19

ЦиньЛисунQinLisong

Цинь Лисун, аспирант

620062, Екатеринбург, ул. Мира, д. 19

Qin Lisong, graduate student

620062, Yekaterinburg, Mira st., 19

Уральский федеральный университет имени первого президента России Б. Н. ЕльцинаРоссияUral Federal University named after the first President of Russia B. N. YeltsinRussian Federation20260205202601176194Copyright © Международный издательский дом научной периодики "Спейс, 20262026Международный издательский дом научной периодики "СпейсМеждународный издательский дом научной периодики "Спейсhttps://www.isjaee.com/jour/about/submissions#copyrightNoticehttps://www.isjaee.com/jour/article/view/2772

В работе рассматривается задача суточного многократного прогнозирования электрической нагрузки при дискретизации 15 минут (24 ч, 96 шагов), где нелинейность, нестационарность и многомасштабное смешение осложняют точное долгогоризонтное предсказание. Предлагается схема EMD-двухветвевой MLP с покомпонентным моделированием и калиброванным слиянием. В частности, эмпирическая модовая декомпозиция (EMD) применяется для разложения ряда нагрузки и последующей реконструкции в высокочастотную (HFC) и низкочастотную (LFC) компоненты; для HFC и LFC обучаются две прямые многовыходные MLP-модели с корреляционным отбором признаков (Пирсон), а для компенсации смещения реконструкции используется глобальное аффинное слияние (Global Affine) в качестве пост-калибровки. Многоцелевая оптимизация гиперпараметров выполняется алгоритмом NSGA-II (Optuna) путём совместной минимизации валидационных ошибок ветвей HFC и LFC. Эксперименты на наборе UCI ElectricityLoadDiagrams (MT_232, 2012-2013) с температурными рядами IPMA (Лиссабон) показывают, что EMD_MLP достигает RMSE/MAE/sMAPE = 8.567/5.434/7.546, снижая RMSE на 16,6 % и 19,5 % по сравнению с MLP и LSTM; вариант с NSGA-II дополнительно улучшает результаты до 8.363/5.377/7.503 и обеспечивает более стабильный профиль ошибки по горизонту. Предложенный метод формирует надёжные высокочастотные суточные траектории нагрузки для задач оперативного планирования, включая unit commitment, резервирование и рыночные операции.

This study addresses day-ahead multi-step load forecasting at 15-min resolution (24 h, 96 steps), where strong nonlinearity, non-stationarity and multi-scale mixing hinder accurate long-horizon prediction. We propose an EMD-dual-branch MLP framework with component-wise modeling and calibrated fusion. Specifically, the original load series is decomposed by empirical mode decomposition and reconstructed into a high-frequency component (HFC) and a low-frequency component (LFC) using a predictability-oriented split criterion. Two direct multi-output multilayer perceptron (MLP) predictors are trained separately for the HFC and LFC branches, with branch-wise Pearson-based feature selection applied to reduce redundancy and enhance interpretability. To mitigate reconstruction bias, a Global Affine fusion is applied as post-hoc calibration to mitigate reconstruction bias. Multi-objective hyperparameter optimization is performed via NSGA-II (Optuna) by jointly minimizing validation errors of the HFC and LFC branches. Experiments on UCI ElectricityLoadDiagrams (MT_232, 2012–2013) with IPMA temperature series (Lisbon) show that EMD_MLP achieves RMSE/MAE/sMAPE = 8.567/5.434/7.546, reducing RMSE by 16.6 % and 19.5 % versus MLP and LSTM; the NSGA-II variant further improves to 8.363/5.377/7.503 and yields a more stable horizon-wise error profile. The proposed method provides reliable high-resolution day-ahead load trajectories for operational scheduling such as unit commitment, reserve scheduling and market operation.

суточный прогноз нагрузкиэмпирическая модовая декомпозициямноговыходной многослойный перцептронNSGA-IIустойчивость по горизонтуday-ahead forecastingempirical mode decompositiondirect multi-output multilayer perceptronNSGA-IImulti-horizon stabilityReferencesKondaiah V., Saravanan B., Sanjeevikumar P. et al. A review on short‐term load forecasting models for micro‐grid application // The Journal of Engineering. – 2022; 2022(7):665-689. Doi: 10.1049/tje2.12151Kondaiah V., Saravanan B., Sanjeevikumar P. et al. A review on short‐term load forecasting models for micro‐grid application // The Journal of Engineering. – 2022; 2022(7):665-689. Doi: 10.1049/tje2.12151Akhtar S., Shahzad S., Zaheer A. et al. Short-Term Load Forecasting Models: A Review of Challenges, Progress, and the Road Ahead // Energies. – 2023; 16(10):4060. Doi: 10.3390/en16104060Akhtar S., Shahzad S., Zaheer A. et al. Short-Term Load Forecasting Models: A Review of Challenges, Progress, and the Road Ahead // Energies. – 2023; 16(10):4060. Doi: 10.3390/en16104060Hu Y., Ye C., Ding Y. et al. Short-term load forecasting utilizing wavelet transform and time series considering accuracy feedback // Int Trans Electr Energy Syst. – 2020. Doi: 10.1002/2050-7038.12455.Hu Y., Ye C., Ding Y. et al. Short-term load forecasting utilizing wavelet transform and time series considering accuracy feedback // Int Trans Electr Energy Syst. – 2020. Doi: 10.1002/2050-7038.12455.Eren Y., Küçükdemiral İ. A comprehensive review on deep learning approaches for short-term load forecasting // Renewable and Sustainable Energy Reviews. – 2024(189); Part B:114031 Doi: 10.1016/j.rser.2023.114031.Eren Y., Küçükdemiral İ. A comprehensive review on deep learning approaches for short-term load forecasting // Renewable and Sustainable Energy Reviews. – 2024(189); Part B:114031 Doi: 10.1016/j.rser.2023.114031.Torkzadeh R., Mirzaei A., Mirjalili M. et al. Medium term load forecasting in distribution systems based on multi linear regression & principal component analysis: A novel approach. 2014 19th Conference on Electrical Power Distribution Networks (EPDC). – 2014: 66-70. Doi: 10.1109/EPDC.2014.6867500.Torkzadeh R., Mirzaei A., Mirjalili M. et al. Medium term load forecasting in distribution systems based on multi linear regression & principal component analysis: A novel approach. 2014 19th Conference on Electrical Power Distribution Networks (EPDC). – 2014: 66-70. Doi: 10.1109/EPDC.2014.6867500.Pandey A., Sahay K., Tripathi M. et al. Short-term load forecasting of UPPCL using ANN. 2014 6th IEEE Power India International Conference (PIICON). – 2014: 1-6. Doi: 10.1109/POWERI.2014.7117741.Pandey A., Sahay K., Tripathi M. et al. Short-term load forecasting of UPPCL using ANN. 2014 6th IEEE Power India International Conference (PIICON). – 2014: 1-6. Doi: 10.1109/POWERI.2014.7117741.Aguilar E., Antonio N. Short-Term Electricity Load Forecasting with Machine Learning // Information. – 2021; 12(2):50. Doi:10.3390/info12020050Aguilar E., Antonio N. Short-Term Electricity Load Forecasting with Machine Learning // Information. – 2021; 12(2):50. Doi:10.3390/info12020050Nury H., Hasan K., Alam B. Comparative study of wavelet-ARIMA and wavelet-ANN models for temperature time series data in northeastern Bangladesh // Journal of King Saud University-Science. – 2017; 29(1): 47-61. Doi: 10.1016/j.jksus.2015.12.002Nury H., Hasan K., Alam B. Comparative study of wavelet-ARIMA and wavelet-ANN models for temperature time series data in northeastern Bangladesh // Journal of King Saud University-Science. – 2017; 29(1): 47-61. Doi: 10.1016/j.jksus.2015.12.002Rodrigues F., Pereira F. Beyond Expectation: Deep Joint Mean and Quantile Regression for Spatiotemporal Problems. Transactions on Neural Networks and Learning Systems // IEEE, 2020, 31(12):5377-5389. Doi: 10.1109/TNNLS.2020.2966745.Rodrigues F., Pereira F. Beyond Expectation: Deep Joint Mean and Quantile Regression for Spatiotemporal Problems. Transactions on Neural Networks and Learning Systems // IEEE, 2020, 31(12):5377-5389. Doi: 10.1109/TNNLS.2020.2966745.Liu Y., Zhao J., Liu J. et al. Regional midterm electricity demand forecasting based on economic, weather, holiday, and events factors. IEEJ Transactions on Electrical and Electronic Engineering. – 2020; 15(2):225-234. Doi: 10.1002/tee.23049Liu Y., Zhao J., Liu J. et al. Regional midterm electricity demand forecasting based on economic, weather, holiday, and events factors. IEEJ Transactions on Electrical and Electronic Engineering. – 2020; 15(2):225-234. Doi: 10.1002/tee.23049Sun G., Jiang C., Wang X. et al. Short‐term building load forecast based on a data‐mining feature selection and LSTM‐RNN method // IEEJ Transactions on Electrical and Electronic Engineering. – 2020; 15(7):1002-1010. Doi: 10.1002/tee.23144Sun G., Jiang C., Wang X. et al. Short‐term building load forecast based on a data‐mining feature selection and LSTM‐RNN method // IEEJ Transactions on Electrical and Electronic Engineering. – 2020; 15(7):1002-1010. Doi: 10.1002/tee.23144Kong W., Dong Y., Jia Y. et al. Short-Term Residential Load Forecasting Based on LSTM Recurrent Neural Network. Transactions on Smart Grid // IEEE. – 2019; 10(1):841-851. Doi: 10.1109/TSG.2017.2753802.Kong W., Dong Y., Jia Y. et al. Short-Term Residential Load Forecasting Based on LSTM Recurrent Neural Network. Transactions on Smart Grid // IEEE. – 2019; 10(1):841-851. Doi: 10.1109/TSG.2017.2753802.Yang L., Wu H., Ding M. et al. Short-term load forecasting in renewable energy grid based on bi-directional long short-term memory network considering feature selection // Automation of Electric Power Systems. – 2021; 45(3):166-173. Doi: 10.7500/AEPS20200202002Yang L., Wu H., Ding M. et al. Short-term load forecasting in renewable energy grid based on bi-directional long short-term memory network considering feature selection // Automation of Electric Power Systems. – 2021; 45(3):166-173. Doi: 10.7500/AEPS20200202002Morais S., Aquila G., de Faria, D. et al. Short-term load forecasting using neural networks and global climate models: An application to a large-scale electrical power system // Applied Energy. – 2023; 348:121439. Doi: 10.1016/j.apenergy.2023.121439.Morais S., Aquila G., de Faria, D. et al. Short-term load forecasting using neural networks and global climate models: An application to a large-scale electrical power system // Applied Energy. – 2023; 348:121439. Doi: 10.1016/j.apenergy.2023.121439.Ahranjani K., Beiraghi M., Ghanizadeh R. Short time load forecasting for Urmia city using the novel CNN-LTSM deep learning structure // Electr Eng. – 2025; 107:1253-1264. Doi: 10.1007/s00202-024-02361-4Ahranjani K., Beiraghi M., Ghanizadeh R. Short time load forecasting for Urmia city using the novel CNN-LTSM deep learning structure // Electr Eng. – 2025; 107:1253-1264. Doi: 10.1007/s00202-024-02361-4Shuang R., Kai Y., Jicai S. et al. Short-term Power Load Forecasting Based on CNN-BiGRU-Attention // J. Electr. Eng. – 2024; 19:344-350. Doi: 10.11985/2024.01.037Shuang R., Kai Y., Jicai S. et al. Short-term Power Load Forecasting Based on CNN-BiGRU-Attention // J. Electr. Eng. – 2024; 19:344-350. Doi: 10.11985/2024.01.037Mounir N., Ouadi H, Jrhilifa I. Short-term electric load forecasting using an EMD-BI-LSTM approach for smart grid energy management system // Energy and Buildings. – 2023; 288:113022. Doi: 10.1016/j.enbuild.2023.113022.Mounir N., Ouadi H, Jrhilifa I. Short-term electric load forecasting using an EMD-BI-LSTM approach for smart grid energy management system // Energy and Buildings. – 2023; 288:113022. Doi: 10.1016/j.enbuild.2023.113022.Zhou Y., Su Z., Gao K. et al. A short-term electricity load forecasting method integrating empirical modal decomposition with SAM-LSTM // Frontiers in Energy Research. – 2024; 12:1423692. Doi: 10.3389/fenrg.2024.1423692.Zhou Y., Su Z., Gao K. et al. A short-term electricity load forecasting method integrating empirical modal decomposition with SAM-LSTM // Frontiers in Energy Research. – 2024; 12:1423692. Doi: 10.3389/fenrg.2024.1423692.Su Z., Zeng J., Liu J. A Short-Term Power Load Forecasting Approach Based on EMD-SSA-LSTM // Journal of Physics: Conference Series. – 2024; 2774-012005. Doi: 10.1088/1742-6596/2774/1/012005.Su Z., Zeng J., Liu J. A Short-Term Power Load Forecasting Approach Based on EMD-SSA-LSTM // Journal of Physics: Conference Series. – 2024; 2774-012005. Doi: 10.1088/1742-6596/2774/1/012005.Lotfipoor A., Patidar S., Jenkins P. Deep neural network with empirical mode decomposition and Bayesian optimisation for residential load forecasting // Expert Systems with Applications. – 2024; 237: Part A, 121355. Doi: 10.1016/j.eswa.2023.121355.Lotfipoor A., Patidar S., Jenkins P. Deep neural network with empirical mode decomposition and Bayesian optimisation for residential load forecasting // Expert Systems with Applications. – 2024; 237: Part A, 121355. Doi: 10.1016/j.eswa.2023.121355.Alberg D., Last M. Short-term load forecasting in smart meters with sliding window-based ARIMA algorithms // Vietnam J Comput Sci. – 2018; 5:241-249. Doi: 10.1007/s40595-018-0119-7Alberg D., Last M. Short-term load forecasting in smart meters with sliding window-based ARIMA algorithms // Vietnam J Comput Sci. – 2018; 5:241-249. Doi: 10.1007/s40595-018-0119-7Zheng Z., Chen H., Luo X. A Kalman filter-based bottom-up approach for household short-term load forecast // Applied Energy. – 2019; 250:882-894. Doi: 10.1016/j.apenergy.2019.05.102Zheng Z., Chen H., Luo X. A Kalman filter-based bottom-up approach for household short-term load forecast // Applied Energy. – 2019; 250:882-894. Doi: 10.1016/j.apenergy.2019.05.102Wang Y., Sun S., Chen X. et al. Short-term load forecasting of industrial customers based on SVMD and XGBoost // International journal of electrical power & energy systems. – 2021; 129:106830. Doi: 10.1016/j.ijepes.2021.106830Wang Y., Sun S., Chen X. et al. Short-term load forecasting of industrial customers based on SVMD and XGBoost // International journal of electrical power & energy systems. – 2021; 129:106830. Doi: 10.1016/j.ijepes.2021.106830Zhang L., Jánošík D. Enhanced short-term load forecasting with hybrid machine learning models: CatBoost and XGBoost approaches // Expert systems with applications. – 2024; 241:122686. Doi: 10.1016/j.eswa.2023.122686Zhang L., Jánošík D. Enhanced short-term load forecasting with hybrid machine learning models: CatBoost and XGBoost approaches // Expert systems with applications. – 2024; 241:122686. Doi: 10.1016/j.eswa.2023.122686Li K., Huang W., Hu G. et al. Ultra-short term power load forecasting based on CEEMDAN-SE and LSTM neural network // Energy and Buildings. – 2023; 279:112666. Doi: 10.1016/j.enbuild.2022.112666Li K., Huang W., Hu G. et al. Ultra-short term power load forecasting based on CEEMDAN-SE and LSTM neural network // Energy and Buildings. – 2023; 279:112666. Doi: 10.1016/j.enbuild.2022.112666Ran P., Dong K., Liu X. et al. Short-term load forecasting based on CEEMDAN and Transformer // Electric power systems research. – 2023; 214: part A, 108885. Doi: 10.1016/j.epsr.2022.108885Ran P., Dong K., Liu X. et al. Short-term load forecasting based on CEEMDAN and Transformer // Electric power systems research. – 2023; 214: part A, 108885. Doi: 10.1016/j.epsr.2022.108885Luo H., Zhang H., Wang J. Ensemble power load forecasting based on competitive-inhibition selection strategy and deep learning // Sustainable energy technologies and assessments. 2022, 51: 101940. Doi: 10.1016/j.seta.2021.101940Luo H., Zhang H., Wang J. Ensemble power load forecasting based on competitive-inhibition selection strategy and deep learning // Sustainable energy technologies and assessments. 2022, 51: 101940. Doi: 10.1016/j.seta.2021.101940Cui J., Kuang W., Geng K. et al. Advanced Short-Term Load Forecasting with XGBoost-RF Feature Selection and CNN-GRU // Processes. – 2024; 12(11):2466. Doi: 10.3390/pr12112466Cui J., Kuang W., Geng K. et al. Advanced Short-Term Load Forecasting with XGBoost-RF Feature Selection and CNN-GRU // Processes. – 2024; 12(11):2466. Doi: 10.3390/pr12112466Asiri M., Aldehim G., Alotaibi F. et al. Short-Term Load Forecasting in Smart Grids Using Hybrid Deep Learning in IEEE Access. – 2024; 12:23504-23513. Doi: 10.1109/ACCESS.2024.3358182.Asiri M., Aldehim G., Alotaibi F. et al. Short-Term Load Forecasting in Smart Grids Using Hybrid Deep Learning in IEEE Access. – 2024; 12:23504-23513. Doi: 10.1109/ACCESS.2024.3358182.Trindade A. ElectricityLoadDiagrams20112014. UCI Machine Learning Repository [Data set]. – 2015. https://doi.org/10.24432/C58C86.Trindade A. ElectricityLoadDiagrams20112014. UCI Machine Learning Repository [Data set]. – 2015. https://doi.org/10.24432/C58C86.IPMA, n. d. Long data series (Séries longas): Lisboa/Geofísico station (daily Tmin, Tmean, Tmax) [Data set]. Instituto Português do Mar e da Atmosfera. Accessed 12 Feb 2026. https://www.ipma.pt/en/oclima/series.longas/?loc=Lisboa%2FGeof%C3%ADsico&type=raw&utm_source=chatgpt.comIPMA, n. d. Long data series (Séries longas): Lisboa/Geofísico station (daily Tmin, Tmean, Tmax) [Data set]. Instituto Português do Mar e da Atmosfera. Accessed 12 Feb 2026. https://www.ipma.pt/en/oclima/series.longas/?loc=Lisboa%2FGeof%C3%ADsico&type=raw&utm_source=chatgpt.comAl-Bossly A. Metaheuristic Optimization with Deep Learning Enabled Smart Grid Stability Prediction // Computers, Materials & Continua. – 2023; 75(3):6395-6408. Doi:10.32604/cmc.2023.028433Al-Bossly A. Metaheuristic Optimization with Deep Learning Enabled Smart Grid Stability Prediction // Computers, Materials & Continua. – 2023; 75(3):6395-6408. Doi:10.32604/cmc.2023.028433Huang E., Shen Z., Long R. et al. The empirical model decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis // Proceedings of the Royal Society of London Series A: mathematical, physical and engineering sciences. – 1998; 454(1971): 903-995. Doi: 10.1098/rspa.1998.0193Huang E., Shen Z., Long R. et al. The empirical model decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis // Proceedings of the Royal Society of London Series A: mathematical, physical and engineering sciences. – 1998; 454(1971): 903-995. Doi: 10.1098/rspa.1998.0193Yin C., Wei N., Wu J. et al. An Empirical Mode Decomposition-Based Hybrid Model for Sub-Hourly Load Forecasting // Energies. – 2024; 17(2):307. Doi: 10.3390/en17020307Yin C., Wei N., Wu J. et al. An Empirical Mode Decomposition-Based Hybrid Model for Sub-Hourly Load Forecasting // Energies. – 2024; 17(2):307. Doi: 10.3390/en17020307Tian F., Fan X., Wang R. et al. A power forecasting method for ultra-short-term photovoltaic power generation using transformer model // Math. Probl. Eng. – 2022 (2022). Doi: 10.1155/2022/9421400Tian F., Fan X., Wang R. et al. A power forecasting method for ultra-short-term photovoltaic power generation using transformer model // Math. Probl. Eng. – 2022 (2022). Doi: 10.1155/2022/9421400Roy S., Mehera R., Pal K. et al. Hyperparameter optimization for deep neural network models: a comprehensive study on methods and techniques // Innovations Syst Softw Eng. – 2025; 21:789-800. Doi: 10.1007/s11334-023-00540-3.Roy S., Mehera R., Pal K. et al. Hyperparameter optimization for deep neural network models: a comprehensive study on methods and techniques // Innovations Syst Softw Eng. – 2025; 21:789-800. Doi: 10.1007/s11334-023-00540-3.Deb K., Pratap A., Agarwal S. et al. A fast and elitist multiobjective genetic algorithm: NSGA-II. IEEE Transactions on Evolutionary Computation. – 2002; 6:182-197. Doi: 10.1109/4235.996017Deb K., Pratap A., Agarwal S. et al. A fast and elitist multiobjective genetic algorithm: NSGA-II. IEEE Transactions on Evolutionary Computation. – 2002; 6:182-197. Doi: 10.1109/4235.996017Ibrahim M., Harkouss F., Biwole P. et al. Multi-objective hyperparameter optimization of artificial neural network in emulating building energy simulation // Energy and Buildings. 2025; 337:115643. Doi: 10.1016/j.enbuild.2025.115643Ibrahim M., Harkouss F., Biwole P. et al. Multi-objective hyperparameter optimization of artificial neural network in emulating building energy simulation // Energy and Buildings. 2025; 337:115643. Doi: 10.1016/j.enbuild.2025.115643Nti K., Teimeh M., Nyarko-Boateng O. et al. Electricity load forecasting: a systematic review // Journal of Electrical Systems and Inf Technol. – 2020; 7(13). Doi: 10.1186/s43067-020-00021-8.Nti K., Teimeh M., Nyarko-Boateng O. et al. Electricity load forecasting: a systematic review // Journal of Electrical Systems and Inf Technol. – 2020; 7(13). Doi: 10.1186/s43067-020-00021-8.Makridakis S., Spiliotis E., Assimakopoulos V. The M4 Competition: Results, findings, conclusion and way forward // International Journal of Forecasting. – 2018; 34(4):802-808. Doi: 10.1016/j.ijforecast.2018.06.001Makridakis S., Spiliotis E., Assimakopoulos V. The M4 Competition: Results, findings, conclusion and way forward // International Journal of Forecasting. – 2018; 34(4):802-808. Doi: 10.1016/j.ijforecast.2018.06.001

The authors declare that there are no conflicts of interest present.