Short-Term Load Forecasting Using Modwt-Lstm: MultiScale Modeling With Temporal And Statistical Analysis
IOSR Journal of Electrical and Electronics Engineering · 2026-02 · 21 (2) · 10-18
DOI: 10.9790/0853-2102011018 ↗Kurz gefasst
Fachartikel in der Zeitschrift IOSR Journal of Electrical and Electronics Engineering, 2026-02. Verfasst von Do Quynh Huong und 1 weiteren Autor:innen.
Abstract (Originalsprache)
Short-term load forecasting is a critical component in the operation and management of modern power systems, particularly in microgrid environments with high variability. This study proposes a hybrid forecasting framework based on the Maximal Overlap Discrete Wavelet Transform and Long Short-Term Memory networks (MODWTLSTM) for hour-ahead electricity demand prediction using a Japan dataset. The MODWT is employed to decompose the original load signal into multiple frequency components, enabling the LSTM model to capture both high-frequency fluctuations and long-term temporal dependencies effectively. To validate the robustness of the proposed model, extensive experiments are conducted against multiple baseline models, including RNN, LSTM, GRU, CNN-LSTM, CNN-GRU, and DWT-LSTM. Performance is evaluated using MAE, RMSE, and MAPE metrics. Furthermore, the analysis is extended to different temporal scenarios, including weekday and specialday (holiday) conditions, to examine model adaptability under varying demand patterns. An ablation study is performed to assess the contribution of each component, and the Diebold–Mariano (DM) test is applied to verify the statistical significance of forecasting improvements. Experimental results demonstrate that the proposed MODWT-LSTM model consistently outperforms benchmark models across all evaluation scenarios, achieving superior accuracy and stability. The findings highlight the effectiveness of integrating wavelet-based decomposition with deep learning for enhancing short-term load forecasting performance in real-world energy systems.
Zitieren
APA
Huong, D. Q., & Linh, D. P. (2026). Short-Term Load Forecasting Using Modwt-Lstm: MultiScale Modeling With Temporal And Statistical Analysis. IOSR Journal of Electrical and Electronics Engineering, 21(2), 10-18. https://doi.org/10.9790/0853-2102011018
Vancouver
Huong DQ, Linh DP. Short-Term Load Forecasting Using Modwt-Lstm: MultiScale Modeling With Temporal And Statistical Analysis. IOSR Journal of Electrical and Electronics Engineering. 2026;21(2):10-18. doi:10.9790/0853-2102011018
BibTeX
@article{Huong2026,
author = {Huong, Do Quynh and Linh, Do Phuong},
title = {Short-Term Load Forecasting Using Modwt-Lstm: MultiScale Modeling With Temporal And Statistical Analysis},
journal = {IOSR Journal of Electrical and Electronics Engineering},
year = {2026},
volume = {21},
number = {2},
pages = {10--18},
doi = {10.9790/0853-2102011018}
}
EUVIMED