Evaluation of Probability Distribution Models for Seasonal Wind Speed Considering Elevation Variability

This study analyzes seasonal wind speed data at Kauai, Hawaii, at altitudes of 80m, 100m, and 120m to identify the most suitable probability distributions. Four commonly used distributions - Weibull, Rayleigh, Lognormal, and Generalized Extreme Value (GEV)—were examined. Distribution parameters were estimated using the Maximum Likelihood Estimation (MLE) method, and all modeling and analysis were performed in MATLAB. Statistical criteria employed to evaluate the performance of each distribution included the Kolmogorov-Smirnov (KS) test, which assesses the goodness-of-fit between the distribution and actual data; the Chi-square test, which determines the best frequency-based fit; the Akaike Information Criterion (AIC), which identifies the distribution with the optimal balance between model fit and complexity; and the Root Mean Square Error (RMSE), which indicates the lowest prediction error. Results indicate that the GEV distribution provided the best overall fit across all altitudes, with the lowest KS (0.0474), RMSE (0.0140), and AIC (42229.09) at 100 m. The Weibull distribution also demonstrated good performance, particularly at 80 m and 100 m, offering a balance between modeling accuracy and simplicity. Conversely, the Rayleigh distribution showed moderate performance, while the Lognormal model exhibited significantly inferior results. These findings underscore the importance of selecting appropriate probability distribution models for different altitudes in wind resource assessment and support effective wind energy system planning.