Differential evolution-based mixture distribution models for wind energy potential assessment: A comparative study for coastal regions of China

Abstract

Mixture distributions generally have higher flexibility than single distributions in describing wind speeds. However, the determination of their components is critical. This work evaluates suitable distributions for the wind energy potential of ten sites along the coast of China. Firstly, ten single distributions are compared to obtain high-quality components for the construction of mixture distributions. Secondly, the best four single distributions are identified based on five goodness-of-fit indicators including root mean square error (RMSE), mean absolute error (MAE), chi-square test (X2), coefficient of determination (R2), and mean absolute percentage error (MAPE), and two-by-two combinations are made to construct ten mixture distributions. Finally, these twenty distributions are comprehensively compared and the wind power density is evaluated using the best distributions. In addition, differential evolution is applied to optimize the model parameters. The simulation results show that Burr, three-parameter Weibull, Nakagami, and two-parameter Weibull are the best four single distributions, while all the mixture distributions significantly outperform the single distributions consistently. This indicates that the mixture models have higher flexibility to capture the potential complexity in the wind speeds. In the wind power density calculations, all regions are over 200 W/m2, with Zhangzhou having the highest density and Haikou the lowest.