Model predictive control of wind turbine fatigue via online rainflow-counting on stress history and prediction

Abstract

The standard fatigue estimation procedure is implemented in Model Predictive Control via externalization of the Rainflow algorithm from the optimization problem. Additionally, stress history is considered in a consistent manner by employing a so-called stress residue. The formulation is implemented in the state-of-the-art MPC framework acados and tested in closed-loop with the 5MW onshore turbine in OpenFAST. Simulation results indicate that the new formulation outperforms conventional PID-and MPC-controllers over the entire wind regime, and that the consideration of stress history is highly beneficial.