Copula autoregressive methodology for the simulation of wind speed and direction time series

In this paper we present a methodology for synthetic generation of wind speed and direction bivariate time series based on copula functions to represent the temporal and cross-dependence structure. We explore the advantages of using this nonlinear time series method over more traditional approaches that use a transformation to normal distributions as an intermediate step. The use of copulas gives some flexibility to represent the serial variability of the real data on the simulation, besides allowing more control on the desired properties of the data. Empirical Bernstein copulas were used to consider the circular nature of wind direction. Experimental analysis and real data application prove the usability and convenience of the proposed methodology.