This is an unedited manuscript accepted for publication and provided as an Article in Press for early access at the author’s request. The article will undergo copyediting, typesetting, and galley proof review before final publication. Please be aware that errors may be identified during production that could affect the content. All legal disclaimers of the journal apply.
Samuel Edet Bassey,
- Research Scholar, Department of Mechanical and Aerospace Engineering University of Uyo, Akwa Ibom State, Nigeria
Abstract
This study assesses ARIMA models for capturing, predicting, and visualizing solar irradiance patterns in Uyo, Nigeria. Previous studies specific to Uyo solar irradiance forecasting explored regression models, which are not suitable for advanced analysis. This gap presented a challenge for accurate energy forecasting, grid integration, and planning. This study utilizes a dataset of approximated solar irradiance values, trained and tested on ARIMA for suitability. The methodology comprises the study location and ARIMA model development. Eighteen ARIMA models were evaluated using the grid search method. The study found ARIMA(2,1,2) with a low AIC of 229.556 as the best model for capturing a 24-row, 288-cell solar irradiance observation. The statistical performance analysis reveals that the dataset has a mean ranging from 123 to 192 W/m², a standard deviation between 210 and 300 W/m², and a maximum of 585 to 911 W/m² across the months of January through December, with a total row count of 24 and 288 cells. The Augmented Dickey-Fuller test produced an ADF statistic of -4.457189 with a p-value of 0.000235, while the differenced series yielded an ADF statistic of -2.580365 and a p-value of 0.097124, confirming stationarity of the irradiance series. Forecast values from step 25 to step 29 indicate that irradiance might peak slightly above 2 W/m². Evaluation of eighteen ARIMA models using Akaike Information Criterion showed that ARIMA(2,1,2) achieved the lowest AIC value of 229.556, outperforming competing models such as ARIMA(2,1,1) with 230.906 and ARIMA(2,1,0) with 231.019. The best model forecast values from step 25 to step 29 show a pattern that will be maintained over future predictions, although the curve will be smoother, while ARIMA(1,1,1) indicates a constant forecast with no change. Correlation, distribution, and residual plots confirmed that the data was properly captured. This is suitable for advanced analysis, as the study provides a background for further research using long-term datasets to improve prediction accuracy and support solar energy planning in Uyo, Nigeria.
Keywords: Solar radiation, Auto-Regressive Integrated Moving Average (ARIMA), Solar irradiance forecasting, Time series analysis, Renewable energy, Uyo, Nigeria
References
- Pelkowski J. A physical rationale for generalized Ångström–Prescott regression. Sol Energy. 2009;83(7):955-963. doi:10.1016/j.solener.2008.12.011.
- Paulescu, M. Stefu, N., Calinoiu, D., Paulescu, E., Pop, N., Boata, R. and Mares, O. (2016). Angstrom-Prescott Equation: Physical Basis, Empirical Model and Sensitivity Analysis. Renewable Energy Reviews, 62, 495-506. https://doi.org/10.1016/j.ser.2016.04.12
- Paulescu, M., Badescu, V., Budea, S. and Dumitrescu, A. (2022). What Angstrom-Prescott Equation Tells Us About the Cloud and Clear Sky Climatologies. Theoretical and Applied Climatology, 147, 239-250. https://doi.org/10.1007/s00704-021-03805-8
- Augustine C, Nnabuchi MN. Relationship between global solar radiation and sunshine hours for Calabar, Port Harcourt and Enugu, Nigeria. Int J Phys Sci. 2009;4(4):182-188. doi:10.5897/IJPS.9000481.
- Njau, E.C. (1996). Generalized Derivation of the Angstrom-Prescott Equations. Renewable Energy, 7(1), 105-108
- Liu X, Xu Y, Zhong X, Zhang W, Porter JR, Liu W. Assessing models for parameters of the Ångström–Prescott formula in China. Appl Energy. 2012;96:327-338. doi:10.1016/j.apenergy.2011.12.083.
- Mabasa B, Lysko MD, Tazvinga H, Mulaudzi ST, Zwane N, Moloi SJ. The Ångström–Prescott regression coefficients for six climatic zones in South Africa. 2020;13:5418.
- Almorox J, Benito M, Hontoria C. Estimation of monthly Ångström–Prescott equation coefficients from measured daily data in Toledo, Spain. Renew Energy. 2005;30(6):931-936. doi:10.1016/j.renene.2004.08.002.
- Iradukunda C, Chiteka K. Angstrom-Prescott type models for predicting solar irradiation for different locations in Zimbabwe. Stroj Vestn J Mech Eng. 2023;69(1-2):32-48. doi:10.5545/sv-jme.2022.331.
- Ekanem, I. E., Akpan, U. E. and Utah, E.U. (2016). Analysis of Global Solar Radiation in Uyo, Akwa-Ibom State, Nigeria Using Temperature Models. World Journals of Applied Science and Technology (WoJAST), 8(2), 146-153
- Ibeh GF, Agbo GA. Comparison of empirical and artificial neural network models for the correlation of monthly average global solar radiation with sunshine hours in Minna, Niger State, Nigeria. Int J Phys Sci. 2012;7(8). doi:10.5897/IJPS11.1750.
- Augustine C, Nnabuchi MN. Analysis of some meteorological data for some selected cities in the Eastern and Southern zone of Nigeria. Afr J Environ Sci Technol. 2010;4(2):92-99. doi:10.5897/AJEST09.092.
- Almox, J., Benito, M. and Hontoria, C. (2004). Estimation of Monthly Angstrom-Prescott Equation Coefficients from Measured Daily Date in Toledo, Spain. Renewable Energy, 931-936. https://org/10.1016/j.renene.2004.08.002
- Wan, K. K. W., Tang, H. L., Yang, L and Lam, J. C. (2008). An Analysis of Thermal and Solar Radiation Models Using an Angstrom-Prescott Equation and Artificial Neural Networks. Energy, 33, 1115-1127. https://doi.org/10.1016/j.energy.2008.01.015
International Journal of Energy and Thermal Applications
| Volume | 04 | |
| 02 | ||
| Received | 07/07/2026 | |
| Accepted | 23/09/2026 | |
| Published | 10/10/2026 | |
| Publication Time | 95 Days |