A Multivariate Hybrid ARIMA-GA-SVR Approach for Short-Term Solar Radiation Forecasting

• Andressa F. Arruda UPE
• João F. L. de Oliveira UPE

DOI: https://doi.org/10.5753/eniac.2026.31081

Resumo

Accurate solar radiation forecasting is crucial for integrating photovoltaic energy into the electrical grid and addressing the challenges of intermittency and variability. In this work, a new multivariate hybrid approach was proposed that combines the linear modeling capabilities of ARIMA with the nonlinear learning ability of a Support Vector Machine (SVR). The key to innovation in this method is the design of a multivariate feature set that incorporates past residuals of the target variable, observed values of exogenous meteorological variables, and one-step-ahead linear forecasts for both the target and exogenous variables generated by ARIMA. By including these linear forecasts in the SVR’s input, the model can better capture the dynamic relationship between solar radiation and future meteorological conditions. In addition, a GA-based feature selection step is employed to select the optimal subset of lagged features, aiming to reduce model complexity and computational cost.The proposed ARIMA–GA-SVR model was evaluated using hourly data from five Brazilian cities. Its performance was compared with hybrid models from the literature (OTL-ARIMAX-PSO-MC, ETS-ANN, ARIMA-ANN, KC-MLP) using RMSE, MAE, ARV, and IA as evaluation metrics. The results show that the ARIMA–GA-SVR model achieved the best overall performance across all the analysed cities, outperforming all compared models in most of them. These findings support the conclusion that incorporating linear forecasts of exogenous variables into the nonlinear residual modeling stage is an effective way to improve short-term solar radiation prediction.

Referências

Aguiar, L. M. et al. (2016). Combining solar irradiance measurements, satellite-derived data and a numerical weather prediction model to improve intra-day solar forecasting. Renewable Energy, 97:599–610.

Ahmadi, M., Aly, H., and Khashei, M. (2025). Enhancing power grid stability with a hybrid framework for wind power forecasting: Integrating kalman filtering, deep residual learning, and bidirectional lstm. Energy, 334:137752.

Dong, Z., Yang, D., Reindl, T., and Walsh, W. M. (2015). A novel hybrid approach based on self-organizing maps, support vector regression and particle swarm optimization to forecast solar irradiance. Energy, 82:570–577.

Ghofrani, M., de Rezende, M., Azimi, R., and Ghayekhloo, M. (2016). K-means clustering with a new initialization approach for wind power forecasting. In IEEE Power and Energy Society General Meeting.

Guermoui, M., Melgani, F., Gairaa, K., and Mekhalfi, M. L. (2020). A comprehensive review of hybrid models for solar radiation forecasting. Journal of Cleaner Production, 258:120357.

Hyndman, R. J. and Khandakar, Y. (2008). Automatic time series forecasting: The forecast package for r. Journal of Statistical Software, 27:1–22.

Krishnan, N., Kumar, K. R., and Inda, C. S. (2023). How solar radiation forecasting impacts the utilization of solar energy: A critical review. Journal of Cleaner Production, 388:135860.

Kumari, P. and Toshniwal, D. (2021). Deep learning models for solar irradiance forecasting: A comprehensive review. Journal of Cleaner Production, 318:128566.

Oliveira, J. F. L., Silva, E. G., and de Mattos Neto, P. S. G. (2022). A hybrid system based on dynamic selection for time series forecasting. IEEE Transactions on Neural Networks and Learning Systems, 33(8):3251–3263.

Panigrahi, R. and Behera, A. K. (2017). Exponential smoothing based ann model for time series forecasting. Engineering Applications of Artificial Intelligence, 66:49–63.

Santos, S. et al. (2022). Solar irradiance forecasting using dynamic ensemble selection. Applied Sciences, 12(7):3510.

Solano, E. S., Dehghanian, P., and Affonso, C. M. (2022). Solar radiation forecasting using machine learning and ensemble feature selection. Energies, 15(19):7049.

Wang, Z., Peng, T., Zhang, C., et al. (2024). Enhancing multi-step short-term solar radiation forecasting using hybrid deep learning models. Applied Energy, page 124708.

Zhang, G. P. (2003). Time series forecasting using a hybrid arima and neural network model. Neurocomputing, 50:159–175.

Zhang, H., Liu, X., Gao, X., and Liu, Z. (2025). Hybrid arimax model integrating optimal time-lag selection and markov chain error correction for residential heating load forecasting. Energy Reports, 14:5259–5273.