Wind farm model aggregation using probabilistic clustering

The paper proposes an innovative probabilistic clustering concept for aggregate modeling of wind farms (WFs). The proposed technique determines the number of equivalent turbines that can be used to represent large WF during the year in system studies. Support vector clustering (SVC) technique is used to cluster wind turbines (WTs) based on WF layout and incoming wind. These clusters are then arranged into groups, and finally through analysis of wind at the site, equivalent number of WTs for WF representation is determined. The method is demonstrated on a WF consisting of 49 WTs connected to the grid through two transmission lines. Dynamic responses of the aggregate model of the WF are compared against responses of the full WF model for various wind scenarios.