AI-Driven Predictive Soiling Management for Solar PV Systems: Real-Time Soiling Estimation, Energy-Loss Prediction, and Optimal Cleaning
Main Article Content
Abstract
Dust and particulate accumulation on photovoltaic (PV) module surfaces — commonly termed “soiling” — is one of the most under-managed causes of energy yield loss in utility-scale and distributed solar installations, with reported losses ranging from under 2% to over 30% per month depending on climate, dust load, and cleaning practice. Conventional cleaning strategies rely on fixed calendar schedules or manual visual inspection, which frequently result in either premature, water- and labor-intensive cleaning or delayed cleaning that allows avoidable energy loss to accumulate. This paper proposes an end-to-end, AI-driven predictive soiling management (AI-PSM) framework that (i) estimates the real-time soiling ratio (SR) of PV arrays from low-cost sensor and weather data using a hybrid CNN-LSTM-Attention deep learning model, (ii) forecasts short- and medium-horizon energy-loss trajectories under continued soiling, and (iii) computes an economically optimal, zone-wise cleaning schedule that minimizes the combined cost of energy loss and cleaning operations subject to water-use and labor constraints. The framework integrates low-cost IoT sensing (soiled/reference PV pairs, pyranometry, optical dust sensors, and meteorological inputs), a feature-engineering pipeline informed by domain knowledge of aerosol deposition physics, and a decision layer that couples the predictive model with a mixed-integer/rolling-horizon optimizer. On an illustrative case-study dataset constructed to reflect field-reported soiling dynamics, the proposed hybrid model outperformed classical regression, support vector regression, random forest, and standalone LSTM baselines in soiling-ratio estimation accuracy, and the resulting AI-optimized cleaning schedule reduced total operating cost relative to fixed-interval cleaning while maintaining energy losses within an acceptable threshold. The paper details the system architecture, modeling methodology, evaluation protocol, and economic analysis, and provides a template that can be populated with site-specific field data for deployment and journal submission. The framework is designed to be sensor-agnostic, scalable to multi-site solar farms, and compatible with existing SCADA/monitoring infrastructure
Article Details
Section
How to Cite
References
1. Adkine, N. H., & Jolhe, S. P. (2025). Design of an Intelligent Framework for Dust Typing, Severity Quantification, and Performance Recovery in Photovoltaic Systems Using Multi-Domain Learning and Physics-Guided Reinforcement Control. International Journal of Science and Research (IJSR), 959–974. https://doi.org/10.21275/sr251211080515
2. Agyeman, B. T., Naouri, M., Appels, W. M., Liu, J., & Shah, S. L. (2024). Learning-based multi-agent MPC for irrigation scheduling. Control Engineering Practice, 147, 105908. https://doi.org/10.1016/j.conengprac.2024.105908
3. Bessa, J. G., Micheli, L., Montes‐Romero, J., Almonacid, F., & Férnández, E. F. (2022). Estimation of Photovoltaic Soiling Using Environmental Parameters: A Comparative Analysis of Existing Models. Advanced Sustainable Systems, 6(5). https://doi.org/10.1002/adsu.202100335
4. Cali, Ü., Çatak, F. Ö., & Halden, U. (2024). Trustworthy cyber-physical power systems using AI: dueling algorithms for PMU anomaly detection and cybersecurity. Artificial Intelligence Review, 57(7). https://doi.org/10.1007/s10462-024-10827-x
5. Chadly, A., Aldayyani, H. Y., Hamasha, M. M., Amer, S., Maalouf, M., & Mayyas, A. (2024). Selection of optimal strategy for managing decentralized solar PV systems considering uncertain weather conditions. Scientific Reports, 14(1), 12269. https://doi.org/10.1038/s41598-024-62891-6
6. Conceição, R., González‐Aguilar, J., Merrouni, A. A., & Romero, M. (2022). Soiling effect in solar energy conversion systems: A review. Renewable and Sustainable Energy Reviews, 162, 112434. https://doi.org/10.1016/j.rser.2022.112434
7. Fuchun, X., Yang, C., & Feng, L. (2025). Dual-Layer Meta-Learning Framework for Adaptive Multi-Task Scheduling of Digital Agents in Hotel Operations. Tehnicki Vjesnik - Technical Gazette, 32(5). https://doi.org/10.17559/tv-20250524002692
8. Gurav, R., Massara, E., Song, X., Sinclair, K. F., Brown, E., Kusner, M. J., Poduval, B., & Baydin, A. G. (2024). Probabilistic Forecasting of Radiation Exposure for Spaceflight. In arXiv (Cornell University). Cornell University. https://doi.org/10.48550/arxiv.2411.17703
9. Humairi, A. A., Khalis, E., Hemyari, Z. A. A., & Jung, P. (2025). Machine Learning-Based Predictive Maintenance for Photovoltaic Systems. AI, 6(7), 133. https://doi.org/10.3390/ai6070133
10. Kalimeris, A., Psarros, I., Giannopoulos, G., Terrovitis, M., Papastefanatos, G., & Kotsis, G. (2023). Data-Driven Soiling Detection in PV Modules. IEEE Journal of Photovoltaics, 13(3), 461–466. https://doi.org/10.1109/jphotov.2023.3243719
11. Khan, M. A. Z., Wahab, A., Ali, F., Ahmad, N., Kamran, M. A., & Hassan, A. (2023). Performance of solar roof top panels with disparate particulate accumulation: Exergy analysis on an indoor lab study. PLoS ONE, 18(9). https://doi.org/10.1371/journal.pone.0291018
12. Kishor, I., Mamodiya, U., Patil, V., & Naik, N. (2025). AI-Integrated autonomous robotics for solar panel cleaning and predictive maintenance using drone and ground-based systems. Scientific Reports, 15(1), 32187. https://doi.org/10.1038/s41598-025-17313-6
13. Micheli, L., Férnández, E. F., Aguilera, J., & Almonacid, F. (2020). Economics of seasonal photovoltaic soiling and cleaning optimization scenarios. Energy, 215, 119018. https://doi.org/10.1016/j.energy.2020.119018
14. Micheli, L., Férnández, E. F., Fernández‐Solas, Á., Bessa, J. G., & Almonacid, F. (2021). Analysis and mitigation of nonuniform soiling distribution on utility‐scale photovoltaic systems. Progress in Photovoltaics Research and Applications, 30(3), 211–228. https://doi.org/10.1002/pip.3477
15. Muller, M., Micheli, L., Solas, Á. F., Gostein, M., Robinson, J., Morely, K., Dooraghi, M., Alghamdi, Y. A., Almutairi, Z., Almonacid, F., & Férnández, E. F. (2021). An in‐depth field validation of “DUSST”: A novel low‐maintenance soiling measurement device. Progress in Photovoltaics Research and Applications, 29(8), 953–967. https://doi.org/10.1002/pip.3415
16. Oufadel, A., Ydrissi, M. E., Taabane, I., Azouzoute, A., Aicha, A. H., & Ghennioui, H. (2025). Soft sensors for dust estimation with high accuracy: AI-driven approach for CSP solar mirror soiling classification. International Journal of Sustainable Energy, 44(1). https://doi.org/10.1080/14786451.2025.2475305
17. Pergantis, E. N., Priyadarshan, P., Theeb, N. A., Dhillon, P., Ore, J. P., Ziviani, D., Groll, E. A., & Kircher, K. J. (2024). Field demonstration of predictive heating control for an all-electric house in a cold climate. Applied Energy, 360, 122820. https://doi.org/10.1016/j.apenergy.2024.122820
18. Picotti, G., Moretti, L., Cholette, M. E., Binotti, M., Simonetti, R., Martelli, E., Steinberg, T., & Manzolini, G. (2020). Optimization of cleaning strategies for heliostat fields in solar tower plants. Solar Energy, 204, 501–514. https://doi.org/10.1016/j.solener.2020.04.032
19. Selvi, S. T., Devaraj, V., S, R. P. P., & Subramani, K. (2023). Detection of Soiling on PV Module using Deep Learning. International Journal of Electrical and Electronics Engineering, 10(7), 93–101. https://doi.org/10.14445/23488379/ijeee-v10i7p108
20. Sepúlveda-Oviedo, E. H. (2025). Optimizing PV maintenance: Methods, cleaning frequency, and a selection protocol. Energy Reports, 14, 1578–1605. https://doi.org/10.1016/j.egyr.2025.07.008
21. Shi, H., Yang, D., Wang, W., Fu, D., Gao, L., Zhang, J., Hu, B., Shan, Y., Zhang, Y., Bian, Y., Chen, H., & Xia, X. (2023). First estimation of high-resolution solar photovoltaic resource maps over China with Fengyun-4A satellite and machine learning. OSTI OAI (U.S. Department of Energy Office of Scientific and Technical Information), 184, 113549. https://doi.org/10.1016/j.rser.2023.113549
22. Song, Z., Liu, J., & Yang, H. (2021). Air pollution and soiling implications for solar photovoltaic power generation: A comprehensive review. PolyU Institutional Research Archive (Hong Kong Polytechnic University), 298, 117247. https://doi.org/10.1016/j.apenergy.2021.117247
23. Tahir, M. F., Tzes, A., EL-Fouly, T. H. M., Moursi, M. S. E., & Larik, N. A. (2025). Evaluating soiling effects to optimize solar photovoltaic performance using machine learning algorithms. Energy Conversion and Management X, 26, 100921. https://doi.org/10.1016/j.ecmx.2025.100921
24. Taraglio, S., Chiesa, S., Vito, S. D., Paoloni, M., Piantadosi, G., Zanela, A., & Francia, G. D. (2024). Robots for the Energy Transition: A Review. Processes, 12(9), 1982. https://doi.org/10.3390/pr12091982
25. You, S., Lim, Y. J., Dai, Y., & Wang, C. (2018). On the temporal modelling of solar photovoltaic soiling: Energy and economic impacts in seven cities. Applied Energy, 228, 1136–1146. https://doi.org/10.1016/j.apenergy.2018.07.020
26. Zhang, C., Ma, Y., Mi, Z., Yang, F., & Zhang, L. (2023). A rolling-horizon cleaning recommendation system for dust removal of industrial PV panels. Applied Energy, 353, 122168. https://doi.org/10.1016/j.apenergy.2023.122168
27. Zhou, Z., Xiao, B., Zhang, J., He, Z., Niu, Y., & Yang, L. (2025). Lightweight Multimodal Workflow for Automated Fouling Detection and Decision Making in PV Solar Thermal Power Plants. 1455–1461. https://doi.org/10.1109/ceepe64987.2025.11034130
28. Зайцев, А. А., Кошурников, А. В., Gagarin, V. E., Frolov, D. М., & Rzhanitsyn, G. A. (2025). AI-driven spectral analysis of soil heaving for automated surveys in rail transport infrastructure. AI in Civil Engineering, 4(1). https://doi.org/10.1007/s43503-025-00072-8