Subspace identification of a 5 MW reference wind turbine

Nonlinear systems, a category into which essentially all real-world engineering systems fall, are challenging in their design, response prediction, and control. Simulating their dynamic response requires numerical solvers, and the use of frequency-domain techniques is prohibited, which eliminates many commonly used engineering techniques. Subspace identification is a convenient method to approximate a complex, nonlinear system with a linear model, significantly enhancing simulation time and allowing the use of traditional control schemes. This paper investigates the utility of subspace identification for the National Renewable Energy Laboratory (NREL) 5 MW reference model, which runs on their open-source wind turbine simulator, FAST. The ability of the identified model to predict a variety of outputs is compared. The linear models found with subspace identification are used to create a simple gain-scheduled controller, and its performance is compared with the standard NREL 5 MW controller.

Duke Scholars

Author Henri P. Gavin Civil and Environmental Engineering