Abstract
Accurate forecasting of reservoir levels in hydroelectric power plants is essential for efficient energy generation, operational safety, and sustainable water management. This study proposes a hybrid forecasting framework that integrates seasonal-trend decomposition using loess (STL) with the neural hierarchical interpolation time series (NHITS) model optimized through multi-agent hyperparameter optimization (HPO). The STL filter is employed to remove high-frequency noise and preserve underlying signal trends. NHITS leverages hierarchical multi-scale processing and interpolation-based reconstruction to capture both short- and long-term temporal dependencies, while the multi-agent HPO ensures optimal hyperparameter configuration. The proposed method was evaluated using turbine flow data from the Santo Antônio hydroelectric power plant in Brazil, achieving superior performance compared to state-of-the-art benchmarks across very short- and short-term forecasting horizons.
Data availability
For future comparisons, the dataset is available at: https://dados.ons.org.br/dataset/dados_hidrologicos_ho.
References
Ávila, L. et al. Seasonal streamflow forecast in the Tocantins river basin, Brazil: An evaluation of ECMWF-SEAS5 with multiple conceptual hydrological models. Water15 (9), 1695. https://doi.org/10.3390/w15091695 (2023).
Stefenon, S. F. et al. Neural hierarchical interpolation time series (NHITS) for reservoir level multi-horizon forecasting in hydroelectric power plants. IEEE Access13, 54853–54865. https://doi.org/10.1109/ACCESS.2025.3554446 (2025).
Tripathy, K. P. & Mishra, A. K. Deep learning in hydrology and water resources disciplines: Concepts, methods, applications, and research directions. J. Hydrol.628, 130458. https://doi.org/10.1016/j.jhydrol.2023.130458 (2024).
da Silva, E. C., Finardi, E. C. & Stefenon, S. F. Enhancing hydroelectric inflow prediction in the Brazilian power system: A comparative analysis of machine learning models and hyperparameter optimization for decision support. Electr. Power Syst. Res.230, 110275. https://doi.org/10.1016/j.epsr.2024.110275 (2024).
Larcher, J. H. K., Stefenon, S. F., dos Santos Coelho, L. & Mariani, V. C. Enhanced multi-step streamflow series forecasting using hybrid signal decomposition and optimized reservoir computing models. Expert Syst. Appl.255, 124856. https://doi.org/10.1016/j.eswa.2024.124856 (2024).
Carvalho, J. P. M., Stefenon, S. F., Yow, K.-C., Ovejero, R. G. & Leithardt, V. R. Q. N-beats neural network applied for insulator fault prediction considering emd methods. In International Conference on Electrical Vol. 5 1–6 (IEEE, Zanzibar, Tanzania, 2025).
Dion, P., Martel, J.-L. & Arsenault, R. Hydrological ensemble forecasting using a multi-model framework. J. Hydrol.600, 126537. https://doi.org/10.1016/j.jhydrol.2021.126537 (2021).
Hwan Seo, J. & Kim, K.-D. An rnn-based adaptive hybrid time series forecasting model for driving data prediction. IEEE Access13, 54177–54191. https://doi.org/10.1109/ACCESS.2025.3554803 (2025).
Pranolo, A., Mao, Y., Wibawa, A. P., Utama, A. B. P. & Dwiyanto, F. A. Robust lstm with tuned-pso and bifold-attention mechanism for analyzing multivariate time-series. IEEE Access10, 78423–78434. https://doi.org/10.1109/ACCESS.2022.3193643 (2022).
Gopali, S., Siami-Namini, S., Abri, F. & Siami Namin, A. A comparative multivariate analysis of var and deep learning-based models for forecasting volatile time series data. IEEE Access12, 155423–155436. https://doi.org/10.1109/ACCESS.2024.3447354 (2024).
Yang, Y. et al. Advanced prediction of landslide deformation through temporal fusion transformer and multivariate time-series clustering of insar: Insights from the badui region, eastern tibet. IEEE Trans. Geosci. Remote Sens.62, 1–19. https://doi.org/10.1109/TGRS.2024.3504241 (2024).
Aquino, L. S. et al. Spatiotemporal wind energy forecasting: A comprehensive survey and a deep equilibrium-based case study with StemGNN. IEEE Access13, 131461–131482. https://doi.org/10.1109/ACCESS.2025.3586997 (2025).
Starke, L. et al. Interference recommendation for the pump sizing process in progressive cavity pumps using graph neural networks. Sci. Rep.13 (1), 16884. https://doi.org/10.1038/s41598-023-43972-4 (2023).
Zhao, X. et al. A comprehensive review of methods for hydrological forecasting based on deep learning. Water16 (10), 1407. https://doi.org/10.3390/w16101407 (2024).
Muniz, R. N. et al. Time series forecasting based on multi-criteria optimization for model and filter selection applied to hydroelectric power plants. Energy https://doi.org/10.1016/j.energy.2025.138688 (2025).
dos Santos, G. H. et al. Static attitude determination using convolutional neural networks. Sensors21 (19), 6419. https://doi.org/10.3390/s21196419 (2021).
Stefenon, S. F., Seman, L. O., Aquino, L. S. & dos Santos Coelho, L. Wavelet-seq2seq-lstm with attention for time series forecasting of level of dams in hydroelectric power plants. Energy274, 127350. https://doi.org/10.1016/j.energy.2023.127350 (2023).
Latif, S. D. & Ahmed, A. N. A review of deep learning and machine learning techniques for hydrological inflow forecasting. Environ. Dev. Sustain.25 (11), 12189–12216. https://doi.org/10.1007/s10668-023-03131-1 (2023).
Shen, C., & Lawson, K. Applications of deep learning in hydrology. In Deep Learning for the Earth Sciences: A Comprehensive Approach to Remote Sensing, Climate Science, and Geosciences. 283–297. https://doi.org/10.1002/9781119646181.ch19 (2021).
Corso, M. P. et al. Evaluation of visible contamination on power grid insulators using convolutional neural networks. Electr. Eng.105, 3881–3894. https://doi.org/10.1007/s00202-023-01915-2 (2023).
Stefenon, S. F., Seman, L. O., Pavan, B. A., Ovejero, R. G. & Leithardt, V. R. Q. Optimal design of electrical power distribution grid spacers using finite element method. IET Generat. Trans. Distrib.16 (9), 1865–1876. https://doi.org/10.1049/gtd2.12425 (2022).
Stefenon, S. F., Seman, L. O., Singh, G. & Yow, K.-C. Enhanced insulator fault detection using optimized ensemble of deep learning models based on weighted boxes fusion. Int. J. Electr. Power Energy Syst.168, 110682. https://doi.org/10.1016/j.ijepes.2025.110682 (2025).
Farrokhabadi, M. et al. Day-ahead electricity demand forecasting competition: Post-covid paradigm. IEEE Open Access J. Power Energy9, 185–191. https://doi.org/10.1109/OAJPE.2022.3161101 (2022).
Lee, J., Kang, J., Lee, S. & Oh, H.-M. Ultra-short term photovoltaic generation forecasting based on data decomposition and customized hybrid model architecture. IEEE Access12, 20840–20853. https://doi.org/10.1109/ACCESS.2024.3362234 (2024).
Su, H.-Y. & Huang, C.-R. Enhanced wind generation forecast using robust ensemble learning. IEEE Trans. Smart Grid12 (1), 912–915. https://doi.org/10.1109/TSG.2020.3021578 (2021).
Stefenon, S. F., Singh, G., Souza, B. J., Freire, R. Z. & Yow, K.-C. Optimized hybrid YOLOu-Quasi-ProtoPNet for insulators classification. IET Gener. Trans. Distrib.17(15), 3501–3511. https://doi.org/10.1049/gtd2.12886 (2023).
Lappicy, T. & Lima, C. H. Enhancing monthly streamflow forecasting for brazilian hydropower plants through climate index integration with stochastic methods. RBRH28, 48. https://doi.org/10.1590/2318-0331.282320230118 (2023).
Fredo, G. L. M., Finardi, E. C., Larroyd, P. V. & Picarelli, L. B. Inflow aggregation and run-of-the-river inflow energy for reducing dimensionality in the long-term generation scheduling problem. IEEE Access9, 98542–98560. https://doi.org/10.1109/ACCESS.2021.3096143 (2021).
Barino, F. O., Silva, V. N. H., López-Barbero, A. P., De Mello Honório, L. & Santos, A. B. D. Correlated time-series in multi-day-ahead streamflow forecasting using convolutional networks. IEEE Access8, 215748–215757. https://doi.org/10.1109/ACCESS.2020.3040942 (2020).
Xie, H., Zhang, L. & Lim, C. P. Evolving cnn-lstm models for time series prediction using enhanced grey wolf optimizer. IEEE Access8, 161519–161541. https://doi.org/10.1109/ACCESS.2020.3021527 (2020).
Wan, A., Chang, Q., Khalil, A.-B. & He, J. Short-term power load forecasting for combined heat and power using CNN-LSTM enhanced by attention mechanism. Energy282, 128274. https://doi.org/10.1016/j.energy.2023.128274 (2023).
Rafi, S. H., Deeba, S. R. & Hossain, E. A short-term load forecasting method using integrated CNN and LSTM network. IEEE Access9, 32436–32448. https://doi.org/10.1109/ACCESS.2021.3060654 (2021).
Yamasaki, M. et al. Optimized hybrid ensemble learning approaches applied to very short-term load forecasting. Int. J. Electr. Power Energy Syst.155, 109579. https://doi.org/10.1016/j.ijepes.2023.109579 (2024).
Qin, Q., Huang, Z., Zhou, Z., Chen, Y. & Zhao, W. Hodrick-prescott filter-based hybrid arima–slfns model with residual decomposition scheme for carbon price forecasting. Appl. Soft Comput.119, 108560. https://doi.org/10.1016/j.asoc.2022.108560 (2022).
Li, K., Huang, W., Hu, G. & Li, J. Ultra-short term power load forecasting based on CEEMDAN-SE and LSTM neural network. Energy Build.279, 112666. https://doi.org/10.1016/j.enbuild.2022.112666 (2023).
Stefenon, S. F. & Yow, K.-C. Adaptive filter-driven optimized attention-based cnn-lstm for load forecasting in microgrids. Results Eng. https://doi.org/10.1016/j.rineng.2025.107470 (2025).
Ran, P., Dong, K., Liu, X. & Wang, J. Short-term load forecasting based on CEEMDAN and transformer. Electr. Power Syst. Res.214, 108885. https://doi.org/10.1016/j.epsr.2022.108885 (2023).
Stefenon, S. F. et al. Hypertuned wavelet convolutional neural network with long short-term memory for time series forecasting in hydroelectric power plants. Energy313, 133918. https://doi.org/10.1016/j.energy.2024.133918 (2024).
Klaar, A. C. R., Stefenon, S. F., Seman, L. O., Mariani, V. C. & Coelho, L. S. Structure optimization of ensemble learning methods and seasonal decomposition approaches to energy price forecasting in Latin America: A case study about Mexico. Energies16(7), 3184. https://doi.org/10.3390/en16073184 (2023).
Stefenon, S. F. et al. Group method of data handling using Christiano-Fitzgerald random walk filter for insulator fault prediction. Sensors23(13), 6118. https://doi.org/10.3390/s23136118 (2023).
Seman, L. O., Stefenon, S. F., Mariani, V. C. & Coelho, L. S. Ensemble learning methods using the Hodrick-Prescott filter for fault forecasting in insulators of the electrical power grids. Int. J. Electr. Power Energy Syst.152, 109269. https://doi.org/10.1016/j.ijepes.2023.109269 (2023).
Li, Z., Liu, Y. & Tanaka, G. Multi-reservoir echo state networks with Hodrick-Prescott filter for nonlinear time-series prediction. Appl. Soft Comput.135, 110021. https://doi.org/10.1016/j.asoc.2023.110021 (2023).
Moreno, S. R., Seman, L. O., Stefenon, S. F., dos Santos Coelho, L. & Mariani, V. C. Enhancing wind speed forecasting through synergy of machine learning, singular spectral analysis, and variational mode decomposition. Energy292, 130493. https://doi.org/10.1016/j.energy.2024.130493 (2024).
Ribeiro, M. H. D. M. et al. Variational mode decomposition and bagging extreme learning machine with multi-objective optimization for wind power forecasting. Appl. Intell.54, 3119–3134. https://doi.org/10.1007/s10489-024-05331-2 (2024).
Gomaa, E. et al. Assessment of hybrid machine learning algorithms using trmm rainfall data for daily inflow forecasting in Três Marias reservoir, eastern Brazil. Heliyon https://doi.org/10.1016/j.heliyon.2023.e18819 (2023).
Li, M.-W., Xu, R.-Z., Geng, J., Hong, W.-C. & Li, H. A ship motion forecasting approach based on fourier transform, regularized bi-lstm and chaotic quantum adaptive woa. Ocean Eng.313, 119560 (2024).
Yang, Z.-Y., Cao, X., Xu, R.-Z., Hong, W.-C. & Sun, S.-L. Applications of chaotic quantum adaptive satin bower bird optimizer algorithm in berth-tugboat-quay crane allocation optimization. Expert Syst. Appl.237, 121471 (2024).
Seman, L. O., Stefenon, S. F., Yow, K.-C., dos Santos Coelho, L. & Mariani, V. C. Multi-step short-term solar energy forecasting using Fourier-enhanced BiLSTM and neural additive models. Renew. Energy257, 124738. https://doi.org/10.1016/j.renene.2025.124738 (2026).
Stefenon, S. F., Cristoforetti, M. & Cimatti, A. Conditional diffusion to enhance performance of object detection in unbalanced data engineering drawings. Neural Comput. Appl.37, 20335–20364. https://doi.org/10.1007/s00521-025-11458-1 (2025).
Borgohain, P. L., Phukan, S. & Ahuja, D. R. Downstream channel changes and the likely impacts of flow augmentation by a hydropower project in river Dikrong, India. Int. J. River Basin Manag.17 (1), 25–35. https://doi.org/10.1080/15715124.2018.1439497 (2019).
Seman, L. O. et al. Simultaneously anomaly detection and forecasting for predictive maintenance using a zero-cost differentiable architecture search-based network. Comput. Ind. Eng.208, 111412. https://doi.org/10.1016/j.cie.2025.111412 (2025).
Stefenon, S. F., Seman, L. O., Klaar, A. C. R., Ovejero, R. G. & Leithardt, V. R. Q. Hypertuned-YOLO for interpretable distribution power grid fault location based on EigenCAM. Ain Shams Eng. J.15 (6), 102722. https://doi.org/10.1016/j.asej.2024.102722 (2024).
Li, W. & Jiang, X. Prediction of air pollutant concentrations based on tcn-bilstm-dmattention with stl decomposition. Sci. Rep.13 (1), 4665. https://doi.org/10.1038/s41598-023-31569-w (2023).
Xu, L. et al. Offshore wind speed assessment with statistical and attention-based neural network methods based on stl decomposition. Renew. Energy216, 119097 (2023).
Gong, J., Qu, Z., Zhu, Z. & Xu, H. Parallel timesnet-bilstm model for ultra-short-term photovoltaic power forecasting using stl decomposition and auto-tuning. Energy320, 135286 (2025).
Zeng, H. et al. A novel hybrid stl-transformer-arima architecture for aviation failure events prediction. Reliab. Eng. Syst. Saf.246, 110089. https://doi.org/10.1016/j.ress.2024.110089 (2024).
Yuan, H., Ma, X., Ma, M. & Ma, J. Hybrid framework combining grey system model with gaussian process and stl for co2 emissions forecasting in developed countries. Appl. Energy360, 122824. https://doi.org/10.1016/j.apenergy.2024.122824 (2024).
Kehinde, T. O., Adedokun, O. J., Kareem, M. K., Akpan, J. & Olanrewaju, O. A. Stl-elm: A computationally efficient hybrid approach for predicting high volatility stock market. Intell. Syst. Appl.27, 200564. https://doi.org/10.1016/j.iswa.2025.200564 (2025).
Challu, C. et al. Nhits: Neural hierarchical interpolation for time series forecasting. Proc. AAAI Conf. Artif. Intell.37, 6989–6997 (2023).
Li, X., Liu, Z., Yuan, C. & Liao, Z. Electric parameter prediction of rod pumping units based on n-hits model. PLoS ONE20 (7), 0326973. https://doi.org/10.1371/journal.pone.0326973 (2025).
Apte, M. & Haribhakta, Y. Advancing financial forecasting: a comparative analysis of neural forecasting models n-hits and n-beats. In World Conference on Information Systems for Business Management, pp. 199–212. https://doi.org/10.1007/978-981-96-1747-0_17 (Springer, 2024).
Muqtadir, A., Li, B., Ying, Z., Songsong, C. & Kazmi, S. N. Day-ahead demand response potential prediction in residential buildings with hitskan: A fusion of kolmogorov-arnold networks and n-hits. Energy Build.332, 115455. https://doi.org/10.1016/j.enbuild.2025.115455 (2025).
Umutoni, L. et al. Decoding time: Unraveling the power of n-beats and n-hits vs. lstm for accurate soil moisture prediction. Comput. Electron. Agric.237, 110614. https://doi.org/10.1016/j.compag.2025.110614 (2025).
Olivares, K.G., Challu, C., Garza, F., Canseco, M.M. & Dubrawski, A. NeuralForecast: User friendly state-of-the-art neural forecasting models. PyCon Salt Lake City, Utah, USA https://github.com/Nixtla/neuralforecast (2022).
Stefenon, S. F., Cristoforetti, M. & Cimatti, A. Automatic digitalization of railway interlocking systems engineering drawings based on hybrid machine learning methods. Expert Syst. Appl.281, 127532. https://doi.org/10.1016/j.eswa.2025.127532 (2025).
Madiraju, M.B. & Madiraju, M.S.P. Optimindtune: A multi-agent framework for intelligent hyperparameter optimization. arXiv preprint arXiv:2505.19205, 1–7 (2025). https://doi.org/10.48550/arXiv.2505.19205
Zuege, C. V. et al. Wind speed forecasting approach using conformal prediction and feature importance selection. Int. J. Electr. Power Energy Syst.168, 110700. https://doi.org/10.1016/j.ijepes.2025.110700 (2025).
Matos-Carvalho, J.P., Stefenon, S.F., Leithardt, V.R.Q. & Yow, K.-C. Time series forecasting based on optimized llm for fault prediction in distribution power grid insulators. arXiv preprint arXiv:2502.17341, 1–32 (2025)
Stefenon, S.F., Matos-Carvalho, J.P., Leithardt, V.R. & Yow, K.-C. Cnn-tft explained by shap with multi-head attention weights for time series forecasting. arXiv preprint arXiv:2510.06840, 1–10. https://doi.org/10.48550/arXiv.2510.06840 (2025).
Branco, N. W., Cavalca, M. S. M. & Ovejero, R. G. Bootstrap aggregation with christiano-fitzgerald random walk filter for fault prediction in power systems. Electr. Eng.106 (3), 3657–3670. https://doi.org/10.1007/s00202-023-02146-1 (2024).
Zheng, H. et al. Energy optimisation in cloud datacentres with mc-tide: Mixed channel time-series dense encoder for workload forecasting. Appl. Energy374, 123903. https://doi.org/10.1016/j.apenergy.2024.123903 (2024).
Souto, H.G. Evaluating the efficacy of nhits for forecasting stock realized volatility: A comparative analysis with established models. Computat. Econ. 1–58. https://doi.org/10.1007/s10614-025-10917-0 (2025).
Majeed, T. et al. Deep learning-based streamflow prediction for western Himalayan river basins. Int. J. Syst. Assurance Eng. Manag. https://doi.org/10.1007/s13198-024-02403-x (2024).
Qiu, X. et al. A k-nearest neighbor attentive deep autoregressive network for electricity consumption prediction. Int. J. Mach. Learn. Cybern.15 (4), 1201–1212. https://doi.org/10.1007/s13042-023-01963-x (2024).
Stefenon, S. F., Seman, L. O., da Silva, L. S. A., Mariani, V. C. & dos Santos Coelho, L. Hypertuned temporal fusion transformer for multi-horizon time series forecasting of dam level in hydroelectric power plants. Int. J. Electr. Power Energy Syst.157, 109876. https://doi.org/10.1016/j.ijepes.2024.109876 (2024).
Xu, D.-M. et al. A novel tcn-augmented cnn-lstm architecture for accurate monthly runoff forecasting. Earth Sci. Inf.18 (3), 467. https://doi.org/10.1007/s12145-025-01966-y (2025).
Cui, Y. et al. Informer model with season-aware block for efficient long-term power time series forecasting. Comput. Electr. Eng.119, 109492. https://doi.org/10.1016/j.compeleceng.2024.109492 (2024).
Qu, Z. et al. Integrated energy short-term multivariate load forecasting based on patchtst secondary decoupling reconstruction for progressive layered extraction multi-task learning network. Expert Syst. Appl.269, 126446. https://doi.org/10.1016/j.eswa.2025.126446 (2025).
Hong, J.-T., Bai, Y.-L., Huang, Y.-T. & Chen, Z.-R. Hybrid carbon price forecasting using a deep augmented fedformer model and multimodel optimization piecewise error correction. Expert Syst. Appl.247, 123325. https://doi.org/10.1016/j.eswa.2024.123325 (2024).
Zhao, J. & Marsani, M. F. Flood level prediction model based on kolmogorov-arnold networks: An improved deep learning approach. Theoret. Appl. Climatol.156 (5), 1–19. https://doi.org/10.1007/s00704-025-05470-7 (2025).
He, J., Yuan, K., Zhong, Z. & Sun, Y. Enhancing short-term power load forecasting with a timesnet-crossformer-lstm approach. IEEE Access12, 56774–56788. https://doi.org/10.1109/ACCESS.2024.3383912 (2024).
Funding
This work was supported by the project Self-adaptive platform based on intelligent agents for the optimization and management of operational processes in logistic warehouses (PLAUTON), PID2023-151701OB-C21, funded by MCIN/AEI/10.13039/501100011033/FEDER, EU. Buratto would like to thank UDESC for the financial support for his interuniversity exchange doctorate at the University of Salamanca. The authors would like to thank the Carlos Chagas Filho Foundation for Research Support in the Fluminense Federal University (UFF). Also, the authors would like to thank the Coordination for the Improvement of Higher Education Personnel (CAPES - Brazil) for the scholarship of the second author and by Council for Scientific and Technological Development (CNPq) for the grant of the last author. This study was financed (i) in part by CAPES under the doctoral scholarship number 88887.808258/2023-00, and (ii) by CNPq under grant number 307858/2025-1. The authors would also like to thank researcher Stefano Stefenon for his valuable contributions to this research.
Author information
Authors and Affiliations
Contributions
Rafael Ninno Muniz and William Gouvêa Buratto - wrote the introduction, method, results, and conclusion sections, drew the figures, and computed the experiments. Gabriel Villarrubia Gonzalez, Laio Oriel Seman, Valdeci José Costa, and Ademir Nied - proofread and supervised.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
About this article
Cite this article
Muniz, R.N., Buratto, W.G., Gonzalez, G.V. et al. Optimized hybrid neural hierarchical interpolation time series with STL for flow forecasting in hydroelectric power plants. Sci Rep (2026). https://doi.org/10.1038/s41598-025-34847-x
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-025-34847-x