Behavioral cloning for fatigue-oriented wind turbine optimal predictive control

As wind energy production keeps on increasing, it is necessary to optimize the fatigue/performance trade-off of operating turbines. However, considering fatigue directly in optimal control needs be carefully done because its faithful model does not necessarily fit standard forms commonly required by general purpose solvers. It was recently shown that a signal variance and induced can statistically related. This suggests fatigue-related cost function expressed as non-quadratic function, used an open-loop problem. also such becomes very efficient only for relatively long prediction horizons. makes on-line solution underlying problem computationally demanding undermines real-time applicability associated MPC schemes. In this paper, behavioral learning proposed imitate controller obtained through open-loop, horizon-based optimization. The cloning compared finely tuned quadratic regarding ability reduce fatigue. Preliminary results show enables significant reduction broad range realistic disturbances while being realtime implementable.