Univariate Solar Power Forecasting Using Higher Correlation Component Selection With Maximal Overlap Discrete Wavelet Transform and Multilayer Feed Forward Neural Network
Energy Technology · 2026-11 · 14 (11)
DOI: 10.1002/ente.70717 ↗Kurz gefasst
Fachartikel in der Zeitschrift Energy Technology, online erschienen am 08.11.2026. Verfasst von Shubham Shringi und 2 weiteren Autor:innen. Die Arbeit nennt 51 Literaturangaben.
Abstract (Originalsprache)
This study introduces a hybrid forecasting framework that integrates higher correlation component selection (HCCS) with maximal overlap discrete wavelet transform‐based multiresolution analysis (MODWT‐MRA), and then deploys a multilayer feedforward neural network (MLFFNN) for enhanced accuracy. In contrast to traditional approaches, the proposed framework employs the Pearson correlation coefficient (PCC) to identify the most informative wavelet coefficients, and thus, dimensionality reduction and increased learning performance. PCC is used to select and keep only those decomposition components that have strong correlations with the target output. The methodology begins by gathering and pre‐processing PV and meteorological data, then performing wavelet decomposition using MODWT‐MRA to obtain multi‐scale signal features. This step in HCCS improves the signal‐to‐noise ratio and reduces irrelevant features. This incorporation of HCCS and MODWT‐MRA is a novel contribution, allowing the model to focus on the most informative signal components before forecasting with MLFFNN. The MLFFNN model is trained on this highly correlated, noise‐free dataset to learn complex nonlinear patterns and produce 1‐day‐ahead forecasts. The results demonstrate that the proposed HCCS‐MODWT‐MRA‐MLFFNN model reflects significant improvement of 71.78% in MAE, 77.50% in RMSE, and 7.44% in R 2 compared to existing benchmark models. Such a study has been conducted to deal with the difficulties of managing highly noisy and weakly correlated data.
Zitieren
APA
Shringi, S., Saini, L. M., & Aggarwal, S. K. (2026). Univariate Solar Power Forecasting Using Higher Correlation Component Selection With Maximal Overlap Discrete Wavelet Transform and Multilayer Feed Forward Neural Network. Energy Technology, 14(11). https://doi.org/10.1002/ente.70717
Vancouver
Shringi S, Saini LM, Aggarwal SK. Univariate Solar Power Forecasting Using Higher Correlation Component Selection With Maximal Overlap Discrete Wavelet Transform and Multilayer Feed Forward Neural Network. Energy Technology. 2026;14(11). doi:10.1002/ente.70717
BibTeX
@article{Shringi2026,
author = {Shringi, Shubham and Saini, Lalit Mohan and Aggarwal, Sanjeev Kumar},
title = {Univariate Solar Power Forecasting Using Higher Correlation Component Selection With Maximal Overlap Discrete Wavelet Transform and Multilayer Feed Forward Neural Network},
journal = {Energy Technology},
year = {2026},
volume = {14},
number = {11},
doi = {10.1002/ente.70717}
}
EUVIMED