Performance analysis of metaheuristic optimization algorithms in estimating the parameters of several wind speed distributions

•11 distributions are used to fit the wind speed data of two Brazilian regions.•The application of optimization methods is expanded to non-conventional distributions.•The coefficient of determination as objective function yielded better results.•Metaheuristic optimization algorithms outperformed Maximum Likelihood method.•Gamma Generalized and Extended Generalized Lindley distributions fitted better. For a better use of wind energy, the accurate selection of the wind speed distributions that best represents the regarding wind regime’s characteristics is essential. The Weibull distribution is the most common, but this model is not always the most suitable. Therefore, in order to obtain more reliable information, the evaluation of different distributions becomes necessary. Another crucial step is the estimation of the parameters that govern these distributions because the accuracy of these estimates directly affects the energy generation calculations. In the last few years, different optimization methods have been used for this purpose. However, the applications of these methods are focused on conventional two-parameter distributions, such as Weibull and Lognormal. Futhermore, different authors report that there is a lack of studies that use optimization methods for this purpose. In this paper, four metaheuristic optimization algorithms (MOA)—namely, Migrating Birds Optimization (MBO), Imperialist Competitive Algorithm (ICA), Harmony Search (HS) and Cuckoo Search (CS)—are used to fit 11 distributions in two Brazillian regions. Thus, this work expands the application of the MOA to beyond the conventional distributions and applies, for the first time, the MBO and ICA in estimating the parameters of wind speed distributions, thereby introducing new ways to optimize the use of wind resources. The fits obtained by the MOA were compared with those obtained by the method Maximum Likelihood Estimation (MLE). Gamma Generalized and Extended Generalized Lindley distributions presented the best fits, and the MOA outperformed the MLE because the global score values obtained were smaller.