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