Abstract
Wind energy is becoming one of the most important elements toward the advancement of sustainable energy systems globally. The assessment of wind energy potential is critical to the optimization of resource application and improvement of technologies. This study focuses on fitting fourteen probability density functions (PDFs) to hourly wind speed data collected from six coastal cities of Pakistan: Gwadar, Jiwani, Karachi, Keti Bandar, Ormara, and Pasni, for the year 2023, measured at 10 m and 50 m heights. These selected distributions are Weibull, Rayleigh, Lognormal, Gamma, Normal, Generalized Extreme Value, Logistic, Nakagami, t Location-Scale, Extreme Value, Inverse Gaussian, Chi-Square, Pearson Type III, and Rician. Four goodness-of-fit (GoF) indices are employed to evaluate the performance of these distributions: root mean square error (RMSE), mean absolute error (MAE), coefficient of determination (R(2)), and chi-squared (χ(2)). These metrics give a clear report on each distribution's aptness to emulate the wind speed information. Observed and computed wind power density (WPD) values are also determined to investigate the application of fitted distribution functions for practical purposes. The inspection of the simulation results shows that GEV, Weibull, Nakagami, and Gamma PDFs proved to be the most promising PDFs for describing wind speed data at 10 m, whereas GEV (predominantly), Weibull, Normal, and Logistic PDFs for wind speed data at 50 m. Further investigation revealed that the GEV distribution consistently exhibited better fitting characteristics, followed closely by Weibull, Nakagami, and Gamma distributions, making them highly suitable for characterizing the wind speed and determining wind energy potential. The extensively used Weibull distribution is not always the first choice. Consequently, the results presented in the paper provide fundamental information about the usage of the resource and energy production for Pakistani coastal wind sites.