Wind speed scenario generation based on dependency structure analysis

Scenario generation is of great significance to scenario-based stochastic programming. The higher the accuracy of the scenarios, the closer the solutions of the problem are to real optimal values. One uncertain variable which is significant in planning microgrids and wind farms is wind speed. Complex dynamic behavior, being non-stationary, and the different type of tail dependence of wind speed data complicate its modeling. Tail dependence is an important feature in modeling accuracy and conventional wind speed scenario generation methods such as autoregressive integrated moving average (ARIMA) and multivariate distribution functions are not capable of modeling tail dependence. The current study uses copula functions with the ability to model tail dependence for the generation of wind speed scenarios. First, the structure of the dependence of wind speed data has been investigated using correlation matrix and tail dependence coefficients. Then, according to the dependence structure of the data, the appropriate copula function has been selected. The results show that the scenarios made by the student-T copula function are more accurate than conventional methods such as ARIMA and the normal multivariate distribution function.