التنبؤ بسرعة الرياح الشهرية في منطقة سبها جنوب غرب ليبيا للفترة 2023-1972 باستخدام النموذج الموسمي الهجين SARIMA-GARCH

Abstract

The global energy transition, driven by accelerating climate change, highlights wind power as a key renewable source; however, its inherent temporal volatility poses substantial forecasting challenges. This study develops a hybrid SARIMA-GARCH model to forecast monthly wind speed in Sabha, southwestern Libya, over the 1972–2023 period. The SARIMA component captures seasonal patterns and linear autocorrelation, while the GARCH component addresses volatility clustering and stochastic residual behavior, utilizing monthly data from the Sabha meteorological station. Following stationarity tests and confirmation of the ARCH effect in the SARIMA residuals, the specific specification SARIMA(1,0,1)(1,1,0)₁₂ - GARCH(1,1) was estimated. Out-of-sample evaluation over the independent test period 2022–2023 compared information criteria and standard forecast accuracy metrics. Empirical results demonstrate that the hybrid model significantly outperforms the standalone seasonal model, with information criteria improving by 6% and forecast errors decreasing by 29% during the test window. All coefficients of the hybrid model are statistically significant, confirming stable conditional variance and persistent volatility dynamics. Forecasts for 2024–2027 indicate annual average wind speeds ranging from 9.38 to 9.73 m/s, maintaining the characteristic spring seasonal peak while confidence intervals widen progressively over longer horizons. Based on these findings, we recommend adopting this hybrid model for short-term forecasts (up to 12 months ahead) to effectively support wind energy project operations in southern Libya. Furthermore, the study advocates extending this analytical framework to other regional stations and incorporating additional exogenous climatic predictors in future research.