Faculty of Electronic Engineering, Niš, Serbia

Faculty of Electronic Engineering, Niš, Serbia

Faculty of Electronic Engineering, Niš, Serbia

Faculty of Electronic Engineering, Niš, Serbia

Faculty of Electronic Engineering, Niš, Serbia

The paper is available at the following link: PDF

Abstract:

This study investigates Multivariate Adaptive Regression Splines (MARS) for predicting hourly photovoltaic power output, comparing performance against Long Short-Term Memory (LSTM) networks. Using three years of data from a 156 kW installation in southeastern Serbia, MARS achieved superior accuracy across mean absolute error, root mean square error, and coefficient of determination. LSTM excelled during stable clear-sky conditions with smooth temporal patterns, while MARS handled rapid weather transitions better through piecewise linear basis functions. MARS automatically identified predictive value in snow depth, precipitable water, and relative humidity— variables excluded from LSTM due to weak linear correlation. These findings demonstrate MARS as an effective, interpretable alternative for solar forecasting involving threshold effects and conditional dependencies.

Keywords:

Solar power prediction, MARS, LSTM, machine learning, renewable energy forecasting, feature selection

REFERENCES:

[1] Sivadharshini B, Archana Saxena, G. Saritha, Ahmad Alawady, Sweta Lall, and Dharmesh Dhabliya, "Solar Energy Forecasting Models for Grid Integration and Power Balancing," E3S Web of Proceedings of 10th Small Systems Simulation Symposium, 2026, Niš, Serbia, 12th February – 14th February 2026 Conferences, vol. 540, 2024, doi: 10.1051/e3sconf/202454010022.
[2] Venkatachalam Mohanasundaram and Balamurugan R, "Photovoltaic solar energy prediction using the seasonal-trend decomposition layer and ASOA optimized LSTM neural network model," Scientific Reports, vol. 15, 2025, doi: 10.1038/s41598-025-87625-0.
[3] Putri Nor Liyana Mohamad Radzi, Saad Mekhilef, Noraisyah Mohamed Shah, Muhammad Naveed Akhter, M. Seyedmahmoudian, and A. Stojcevski, "Optimizing solar power forecasting with metaheuristic algorithms and deep learning models for photovoltaic grid connected systems," Scientific Reports, vol. 15, 2025, doi: 10.1038/s41598-025-23822-1.
[4] Mohamed Sayed Ibrahim, S. Gharghory, and H. Kamal, "A hybrid model of CNN and LSTM autoencoder-based short-term PV power generation forecasting," Electrical Engineering, 2024, doi: 10.1007/s00202-023-02220-8.
[5] Samer Rajah, Francisco J. Muñoz, and Alejandro Rodríguez, "An Optimized Hybrid Framework Based on Long-Short Term Memory Neural Networks and Fourier SynchroSqueezed Transform for Photovoltaic Power Forecasting," IEEE Access, vol. 13, 2025, doi: 10.1109/ACCESS.2025.3589131.
[6] Thipwangmek, N., Woradit, K., Suetrong, N., and Promsuk, N. "Feature Selection Approaches for Short-Term Solar Photovoltaic Power Forecasting," 13th International Conference on Renewable Energy Research and Applications (ICRERA),2024, 252–257. doi:10.1109/ICRERA62673.2024.10815326
[7] N. Radivojević, U. Ilić, A. Petrušić, M. Andrejević Stošović, 'Influence of Solar and Weather Parameter Selection on LSTMbased Model Predictions of Solar Power Production', IcETRAN conference, 2025
[8] Jerome H. Friedman, "Multivariate Adaptive Regression Splines," The Annals of Statistics, vol. 19, no. 1, pp. 1-67, 1991.
[9] M. K et al., "Multivariate wind power curve modeling using multivariate adaptive regression splines and regression trees," 2023.
[10] S. Dubey, J. N. Sarvaiya, and B. Seshadri, 'Temperature Dependent Photovoltaic (PV) Efficiency and Its Effect on PV Production in the World – A Review', Energy Procedia, vol. 33, pp. 311–321, 2013.
[11] M. F. Fitzpatrick and S. G. Warren, "Transmission of Solar Radiation by Clouds over Snow and Ice Surfaces. Part II: Cloud Optical Depth and Shortwave Radiative Forcing from Pyranometer Measurements in the Southern Ocean," Journal of Climate, Nov. 2005, doi: 10.1175/jcli3562.1.
[12] N.Y. Hendrikx, Khadija Barhmi, Lennard Visser, Thomas A. de Bruin, Mário P, Albert Ali Salah, and Wilfried van Sark, "All sky imaging-based short-term solar irradiance forecasting with Long Short-Term Memory networks," Solar Energy, vol. 272, 2024, doi: 10.1016/j.solener.2024.112463.
[13] M. K et al., "Multivariate wind power curve modeling using multivariate adaptive regression splines and regression trees," 2023.
[14] M. S. Abed, F. J. Kadhim, J. K. Almusawi, H. Imran, L. F. A. Bernardo, and S. N. Henedy, "Utilizing multivariate adaptive regression splines (MARS) for precise estimation of soil compaction parameters," Applied Sciences, 2023.
[15] Montaser Abdelsattar, Mohamed A. Ismeil, Mohamed M. A. Azim Zayed, Ahmed Abdelmoety, and Ahmed Emad-Eldeen, "Assessing Machine Learning Approaches for Photovoltaic Energy Prediction in Sustainable Energy Systems," IEEE Access, 2024, doi: 10.1109/access.2024.3437191.
[16] J. M. Bright, 'Solcast: Validation of a satellite-derived solar irradiance dataset', Solar Energy, vol. 189, p. 435-449., 2019, doi: 10.1016/j.solener.2019.07.086.
[17] Meryem Bekar Adiguzel and Mehmet Ali Cengiz, "Model selection in multivariate adaptive regressions splines (MARS) using alternative information criteria," Heliyon, vol. 9, 2023, doi: 10.1016/j.heliyon.2023.e19964.
[18] M. Hansen and C. Kooperberg, "Spline adaptation in extended linear models," Statistical Science, 2002.
[19] Afsaneh Mollasalehi and Armin Farhadi, "Solar and Wind Power Forecasting: A Comparative Review of LSTM, Random Forest, and XGBoost Models," arXiv preprint, 2025, doi: 10.48550/arXiv.2509.24059.