Evaluation of Wind Speed Regime Variability and Model Fit Performance Using Multi-Component Weibull Mixture Functions for Energy Potential Estimation

Abstract Wind speed distribution modelling serves as an essential tool to assess wind energy resources while identifying optimal wind turbine locations. Three statistical models, including Bimodal Mixture Weibull (BMW), Modified Bimodal Mixture Weibull (MBMW) and Four-Component Mixture Weibull (FCMW), were evaluated using wind speed data from Abeokuta (Abk) and Lagos (Lag) in Nigeria. The assessment of each model depended on its parameter estimates and statistical fit to find the best representation of wind characteristics. The BMW exhibited a weak fit and failed to capture detailed wind distribution patterns in Abk; the Root Mean Square Error (RMSE) is 0.9878 and the Kolmogorov-Smirnov (KS) statistic achieved 0.3095. The MBMW model demonstrated superior accuracy through its reduced RMSE value of 0.2774, KS value of 0.0354 and shows a close alignment to actual wind data. The RMSE result of MBMW in Lag reached 0.1119, while BMW produced an RMSE of 2.4927. MBMW produced more accurate histogram matches, identified the main wind speed peak at 4.5 m/s in Abk, 8-9 m/s in Lag and discovered additional peaks that BMW failed to detect. FCMW achieved better accuracy levels than MBMW by accurately modelling multiple peaks complex wind patterns with an RMSE of 0.1082, a KS of 0.0234, along with a KS p-value of 0.9991. The Monte Carlo simulation plots validated these results and the research demonstrates that FCMW outperforms MBMW in wind pattern representation, although MBMW improves upon BMW, thus making FCMW the preferred method for wind energy assessments across diverse West African environments.