Feedback-Optimizing Heuristic Model Predictive Control Applied to a Wind Turbine for Suppressing the Tower Vibrations

Abstract

To have accurate control of a wind turbine, especially when facing wind loads, it is vital to model the dynamic behavior of the system accurately and select the controller parameters properly. To mitigate these challenges, here, an optimal model-based controller is utilized using particle swarm optimization to suppress the vibrations of the tower for an NREL 5 MW turbine in the presence of the turbulence wind. For this purpose, first, a discrete state-space model is derived for the wind turbine based on the finite difference method, and the validity of the model structure is verified with its analytical solution. Then, the structure of a model predictive controller is implemented on the model of the wind turbine with optimization of the plant behavior, extracted from state-space form, in a prediction horizon. Next, to prevent the dependence of the proposed controller structure on its parameters, a number of training scenarios including the turbulent wind with different intensities are run in parallel. The cost of each scenario is continually updated based on the tracking error, the control effort and the control command generation time in total run time. Consequently, the best parameters of the controller are adjusted using a stochastic robust optimization algorithm such as particle swarm optimization. Results demonstrate that the proposed approach has an excellent performance in dissipating the tower vibration.