Forecasting Renewable Energy Shares Using ARIMA and a Grey-Box Model: Evidence from Four Developed and Developing Economies
H. C. Iwu
*
Department of Statistics, Federal University of Technology, P. M. B. 1526, Owerri, Nigeria.
D. C. Bartholomew
Department of Statistics, Federal University of Technology, P. M. B. 1526, Owerri, Nigeria.
O. C. Abiahu
Department of Statistics, Federal University of Technology, P. M. B. 1526, Owerri, Nigeria.
K. T. Onyeawugosi
Department of Statistics, Federal University of Technology, P. M. B. 1526, Owerri, Nigeria.
*Author to whom correspondence should be addressed.
Abstract
This study evaluates ARIMA and hybrid Grey-Box models for forecasting the share of renewable energy in total final energy consumption in Angola, Canada, France, and Nigeria. Annual World Bank data from 1990 to 2021 were analysed using exploratory analysis, stationarity testing, autocorrelation diagnostics, model-selection criteria, and forecast-error measures. Country-specific ARIMA models were fitted, and Random Forest models were then applied to lagged ARIMA residuals to capture additional nonlinear structure. The Grey-Box model produced lower reported mean squared error and mean absolute percentage error than the selected ARIMA model for all four countries. Forecasts for 2022–2026 indicated increasing renewable-energy shares in Angola, Canada, and France, while Nigeria showed comparatively little change. These findings suggest that residual-based integration of statistical and machine-learning models may improve short-term forecasting performance across heterogeneous national energy systems. Interpretation should remain cautious because the analysis is based on short annual time series and does not include additional explanatory variables.
Keywords: Renewable energy forecasting, ARIMA, random forest, grey-box model, time-series analysis, residual learning, renewable energy consumption share, energy planning