A Hybrid Deep-Learning Model for Short-Term Renewable Energy Prediction

Abstract Accurate short-term wind-speed and solar irradiance forecasting is essential for renewable energy integration, grid stability, power dispatch, and energy management. However, most existing prediction models struggle to capture the long-range, nonlinear, and nonstationary behaviors of wind speed and solar irradiance. To address this challenge, this study proposes a hybrid framework combining quantum scalar long short-term memory (QsLSTM), neural ordinary differential equations (NODEs), and multihead attention (MHA) to effectively capture the intrinsic variability, noise, nonlinearity, and complexity of wind-speed and solar irradiance data. The QsLSTM module models temporal sequential dependencies, while NODE enhances continuous temporal representation learning and captures nonlinear dynamics. In addition, the MHA mechanism improves contextual temporal dependency extraction. Experiments on multivariate wind-speed datasets at 10-min and hourly intervals demonstrate that the proposed QsLSTM–NODEs–MHA model consistently outperforms benchmark approaches for 1-day and 1-week forecasting. Using hourly data, it achieved RMSE values of 0.2029 and 0.2423, with R 2 scores of 0.9901 and 0.9944, respectively; for 10-min data, RMSE improved to 0.1161 and 0.1578, with R 2 of 0.9975 and 0.9979, respectively. Likewise, for solar irradiance prediction, our proposed model can reach an R 2 of 0.9669. Uncertainty limit prediction is executed, and the results show that the proposed model has a coverage probability near the nominal coverage probability of 0.90 for wind speed, whereas solar irradiance exhibits moderate coverage of 0.8542 under more strongly intermittent and nonstationary conditions. The minimum normalized average interval widths are 0.0201 for wind speed and 0.1666 for solar irradiance data.

Authors

Institutions

Publication Details

Journal
Journal of Energy Engineering
Published
2026-11-17
DOI
https://doi.org/10.1061/jleed9.eyeng-7294
Primary Topic
Energy Load and Power Forecasting
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

A Hybrid Deep-Learning Model for Short-Term Renewable Energy Prediction

Fu‐Kwun Wang, Alebachew Mengistu Worku
3 links
Journal of Energy Engineering
Energy Load and Power Forecasting
article

A Hybrid Deep-Learning Model for Short-Term Renewable Energy Prediction

Fu‐Kwun Wang, Alebachew Mengistu Worku
article en

Abstract

Abstract Accurate short-term wind-speed and solar irradiance forecasting is essential for renewable energy integration, grid stability, power dispatch, and energy management. However, most existing prediction models struggle to capture the long-range, nonlinear, and nonstationary behaviors of wind speed and solar irradiance. To address this challenge, this study proposes a hybrid framework combining quantum scalar long short-term memory (QsLSTM), neural ordinary differential equations (NODEs), and multihead attention (MHA) to effectively capture the intrinsic variability, noise, nonlinearity, and complexity of wind-speed and solar irradiance data. The QsLSTM module models temporal sequential dependencies, while NODE enhances continuous temporal representation learning and captures nonlinear dynamics. In addition, the MHA mechanism improves contextual temporal dependency extraction. Experiments on multivariate wind-speed datasets at 10-min and hourly intervals demonstrate that the proposed QsLSTM–NODEs–MHA model consistently outperforms benchmark approaches for 1-day and 1-week forecasting. Using hourly data, it achieved RMSE values of 0.2029 and 0.2423, with R 2 scores of 0.9901 and 0.9944, respectively; for 10-min data, RMSE improved to 0.1161 and 0.1578, with R 2 of 0.9975 and 0.9979, respectively. Likewise, for solar irradiance prediction, our proposed model can reach an R 2 of 0.9669. Uncertainty limit prediction is executed, and the results show that the proposed model has a coverage probability near the nominal coverage probability of 0.90 for wind speed, whereas solar irradiance exhibits moderate coverage of 0.8542 under more strongly intermittent and nonstationary conditions. The minimum normalized average interval widths are 0.0201 for wind speed and 0.1666 for solar irradiance data.

Journal of Energy EngineeringVol. 153(1)
National Taiwan University of Science and Technology (TW), Bahir Dar University (ET)
Affordable and clean energy, Climate action
Energy Load and Power Forecasting
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon