Generalized bivariate mixture model of directional wind speed in mixed wind climates

Abstract

Probabilistic modeling of the directional wind speed has been an important topic in the fields of wind energy assessment as well as structural wind-resistant design. The mixture model is required for the probabilistic modeling on the directional wind speed as the wind climates often show the mixture nature. For accuracy and robustness concerns in engineering applications, a generalized bivariate mixture (GBM) model of the directional wind speed is developed by using the copula-based bivariate distribution function. In this model, the marginal distributions and connection function for each sub-model can be flexibly defined. In parametrizing the model, the data-driven parametric estimation including a data preprocess and an automatic unsupervised fitting process is presented. The performance of GBM model is verified using two data sets and compared with the offset elliptical normal mixture (OENM) model. The robustness of GBM model is found for its good performance in fitting to the targets with or without smoothing process, and in tackling directional wind speed with the extremely complex probability characteristic.