https://ifp.hal.science/hal-03010303Collet, DavidDavidColletIFPEN - IFP Energies nouvelles - IFPEN - IFP Energies nouvellesGrenoble INP - Institut polytechnique de Grenoble - Grenoble Institute of Technology - UGA - Université Grenoble AlpesAlamir, MazenMazenAlamirGIPSA-MODUS - GIPSA - Modelling and Optimal Decision for Uncertain Systems - GIPSA-PAD - GIPSA Pôle Automatique et Diagnostic - GIPSA-lab - Grenoble Images Parole Signal Automatique - CNRS - Centre National de la Recherche Scientifique - UGA - Université Grenoble Alpes - Grenoble INP - Institut polytechnique de Grenoble - Grenoble Institute of Technology - UGA - Université Grenoble AlpesDi Domenico, DomenicoDomenicoDi DomenicoIFPEN - IFP Energies nouvelles - IFPEN - IFP Energies nouvellesSabiron, GuillaumeGuillaumeSabironIFPEN - IFP Energies nouvelles - IFPEN - IFP Energies nouvellesNon quadratic smooth model of fatigue for optimal fatigue-oriented individual pitch control.HAL CCSD2020fatigue damagecontrolmonitoring[SDE] Environmental SciencesBelli, Catherine2020-11-17 16:19:552023-08-01 13:01:312020-12-03 18:08:06enJournal articleshttps://ifp.hal.science/hal-03010303/document10.1088/1742-6596/1618/2/022004application/pdf1Operation and maintenance cost of wind turbine is intimately related to fatigue damage. However, considering fatigue directly in an optimization needs to be carefully done because its faithful model does not fit standard forms. Recent results have shown that the fatigue damage of a signal estimated using fatigue theory is non-linearly related to its corresponding variance. Therefore, this relationship is used to design a convex cost function approximating fatigue. In this paper, the sensitivity of open-loop optimizations using this fatigue-oriented non-quadratic cost function to the time horizon considered is studied, in terms of both fatigue and computational cost. Moreover, in order to alleviate the undesirable effects of the computational cost, an efficient fixed-point iteration-based way to optimize this fatigue-oriented cost function is presented. Results show that the fatigue-oriented cost function allows to obtain sensitive fatigue reduction compared to optimizations using a family of classical quadratic cost functions, especially for long optimization horizons. It is eventually, shown that the prediction horizon length required for the non-quadratic criterion to be efficient in an MPC makes its use prohibitive in such framework. However, prospects of reaping the benefits of the fatigue-oriented optimization using imitation learning are given.