Humidity-Aware Hybrid Transformer-LSTM Framework for IoT-Enabled Photovoltaic Power Prediction

Ahmed Mohammed; Ranjit Sarban  Singh; Saad Aslam

doi:10.57159/jcmm.5.2.25394

Authors

Ahmed Mohammed Department of Smart Computing and Cyber Resilience, Faculty of Engineering and Technology, Sunway University, Selangor, Malaysia 47500; Department of Computer Science, Faculty of Science, Gombe State University, Gombe, Nigeria 760253 https://orcid.org/0009-0009-1394-1719
Ranjit Sarban Singh Research Center for Human-Machine Collaboration (HUMAC), Faculty of Engineering and Technology, Sunway University, Selangor, Malaysia 47500 https://orcid.org/0000-0003-2140-2538
Saad Aslam Department of Smart Computing and Cyber Resilience, Faculty of Engineering and Technology, Sunway University, Selangor, Malaysia 47500

DOI:

https://doi.org/10.57159/jcmm.5.2.25394

Keywords:

Photovoltaic Power Prediction, IoT-Based PV Monitoring, Humidity-Aware Modeling, Hybrid Transformer-LSTM, Smart Energy Systems

Abstract

Accurate short-term photovoltaic (PV) power forecasting is essential for Internet of Things (IoT)-enabled monitoring, control, and performance assessment of small-scale solar installations. While electrical variables and temperature are widely used in data-driven PV forecasting models, the contribution of ambient humidity remains insufficiently characterized, particularly in persistently humid environments. This study investigates the role of ambient humidity as a contextual environmental feature and evaluates a humidity-aware Hybrid Transformer-LSTM framework for short-term PV power prediction using real-world IoT data collected from multiple photovoltaic panels over a 34-day monitoring period. The proposed hybrid architecture integrates a Transformer-based self-attention mechanism for cross-feature interaction modeling with LSTM-based recurrent learning to capture temporal persistence. Model performance is evaluated against LSTM-only, Transformer-only, Random Forest, and Linear Regression baselines using a strictly time-ordered train-test split, complemented by architectural and feature ablation studies, rolling time-based validation, cross-panel testing, and robustness analysis under input perturbation. Experimental results show that LSTM-based models achieve the highest predictive accuracy on the short-duration dataset, while ambient humidity provides only marginal and context-dependent benefit as a supplementary environmental feature. Transformer-only models perform poorly under data-limited conditions, while the Hybrid Transformer-LSTM achieves competitive accuracy and demonstrates stable behavior under temporal validation, spatial generalization, and sensor noise. These findings highlight that the primary contribution of this study lies in rigorous evaluation and deployment-aware validation rather than absolute accuracy gains, positioning hybrid attention-recurrent architectures as robustness-oriented solutions for IoT-enabled solar PV monitoring systems.

References

International Energy Agency, "Renewables 2023: Analysis and forecast to 2028," tech. rep., IEA Publications, 2022.

S. Mekhilef, R. Saidur, and M. Kamalisarvestani, "Effect of dust, humidity and air velocity on efficiency of photovoltaic cells," Renewable and Sustainable Energy Reviews, vol. 16, no. 5, pp. 2920–2925, 2012.

E. H. M. Ndiaye, A. Ndiaye, M. Faye, D. Gueye, A. Ba, and M. Traore, "Analysis of the impact of irradiance and temperature on photovoltaic production: A statistical and machine learning approach," MethodsX, vol. 15, p. 103716, 2025.

A. H. Arab, B. Taghezouit, K. Abdeladim, and S. Semaoui, "Maximum power output performance modeling of solar photovoltaic modules," Energy Reports, vol. 6, pp. 680–686, 2020.

M. I. Islam, M. S. Jadin, A. A. Mansur, and T. Alharbi, "Electrical performance and degradation analysis of field-aged PV modules in tropical climates: A comparative experimental study," Energy Conversion and Management: X, vol. 24, p. 100719, 2024.

A. Awasthi, A. K. Shukla, M. M. S. R., C. Dondariya, K. Shukla, D. Porwal, and G. Richhariya, "Review on sun tracking technology in solar PV system," Energy Reports, vol. 6, pp. 392–405, 2020.

R. J. Mustafa, M. R. Gomaa, M. Al-Dhaifallah, and H. Rezk, "Environmental impacts on the performance of solar photovoltaic systems," Sustainability, vol. 12, no. 2, p. 608, 2020.

N. B. Sushmi and D. Subbulekshmi, "Real-time ultra short-term irradiance forecasting using a novel R-GRU model for optimizing PV controller dynamics," Results in Engineering, vol. 26, p. 105046, 2025.

N. H. Alombah, J. N. Mungwe, A. Harrison, W. F. Mbasso, and H. B. Fotsin, "Advanced IoT-based monitoring system for real-time photovoltaic performance evaluation: Conception, development and experimental validation," Scientific African, vol. 28, p. e02763, 2025.

F. Aksan, A. Pawlica, V. Suresh, and P. Janik, "A comparative study of machine learning models for PV energy prediction in an energy community," Energies, vol. 18, no. 22, p. 5980, 2025.

R. Asghar, F. R. Fulginei, M. Quercio, and A. Mahrouch, "Artificial neural networks for photovoltaic power forecasting: A review of five promising models," IEEE Access, vol. 12, pp. 90461–90485, 2024.

K. Ferkous, S. Menakh, M. Guermoui, A. Bellaour, B. Bekkar, A. Rabehi, T. F. Agajie, and M. Benghanem, "Optimized solar power forecasting: A multi-decomposition framework using VMD and swarm techniques," AIP Advances, vol. 15, no. 9, 2025.

M. Tradacete-Ágreda, E. Santiso-Gómez, F. J. Rodríguez-Sánchez, P. J. Hueros-Barrios, J. A. Jiménez-Calvo, and C. Santos-Pérez, "High-performance IoT module for real-time control and self-diagnose PV panels under working daylight and dark electroluminescence conditions," Internet of Things, vol. 25, p. 101006, 2023.

R. H. Casanova and A. Conde, "Enhancement of LSTM models based on data pre-processing and optimization of Bayesian hyperparameters for day-ahead photovoltaic generation prediction," Computers & Electrical Engineering, vol. 116, p. 109162, 2024.

E. Lodhi, X. Liu, G. Xiong, M. A. Khan, Z. Lodhi, T. Nawaz, A. Dilawar, S. Tarkoma, and F. Wang, "SmartPV-AIoT: An AIoT-integrated framework for fault diagnosis and remote monitoring in photovoltaic systems," Energy Conversion and Management: X, vol. 27, p. 101117, 2025.

M. E. M. Ismail, "The impact of cooling systems on the efficiency of solar panels across different climates: An analytical study based on climate variability," SSRN Electronic Journal, 2025.

H. A. Kazem and M. T. Chaichan, "The effect of dust accumulation and cleaning methods on PV panels' outcomes based on an experimental study of six locations in northern Oman," Solar Energy, vol. 187, pp. 30–38, 2019.

N. Jannah, M. S. A. Hanifah, T. S. Gunawan, S. A. Zabidi, S. H. Yusoff, and S. N. M. Sapihie, "Comparative analysis of MLP and CNN-LSTM models for solar power generation forecasting," in Proceedings of the IEEE International Conference on Smart Instrumentation, Measurement and Applications (ICSIMA), pp. 284–289, 2023.

C. Chauhan, G. K. Saxena, C. A. Kshirsagar, R. K. Solanki, G. Kumar, and S. B. Goyal, "Optimizing predictive maintenance in industrial IoT networks using machine learning: A comparative study of SVM, DT and ANN," Journal of Computers, Mechanical and Management, vol. 4, no. 5, pp. 8–16, 2025.