A physically interpretable data-driven surrogate model for wake steering

Abstract. Wake steering models for control purposes are typically based on analytical wake descriptions tuned to match experimental or numerical data. This study explores whether a data-driven surrogate model with high degree of physical interpretation can accurately describe the redirected wake. A linear trained large-eddy-simulation data estimates parameters such as deficit, center location and curliness from measurable inflow turbine variables. These then used generate vertical cross-sections at desired downstream locations. In validation considering eight boundary layers ranging neutral stable conditions, far wake's trajectory, curl available power estimated. significant improvement in accuracy is shown benchmark against two models, especially under derated operating conditions atmospheric stratifications. Even though results not directly generalizable all locations types, outcome this encouraging.