To optimize the energy extraction from the wind or to minimize rotor loads and thus increase its lifetime extension, the control is generally performed globally without considering the local aerodynamics of the blade. The indirect measure of the lift, through a limited number of wall pressure sensors, can be an alternative solution to control the power extracted from the wind at the blade scale. This interest is also emphasized knowing that the wind inflow interaction with blade aerodynamics can lead to power loss, load fluctuations and noise generation (see e.g. [1] [2]).
This project deals with the development of control algorithms [3] in a high fidelity simulated environment. It is applied at the level of the blade section, using global aerodynamic sensors, in order to drive the pitch angle of the blade. The objective is to maintain the aerodynamic lift at its highest value, using different solutions of control, taking into account disturbances caused by turbulent inflows. Some control strategies are currently investigated using a numerical model solving Navier-Stokes equations, the ISIS-CFD solver, developed by CNRS and Centrale Nantes and integrated within the FINE/Marine computing suite distributed by Cadence Design Systems. This solver uses the unsteady Reynolds-averaged Navier-Stokes equations and it is based on fully-unstructured (face-based) finite volume discretization [4]. A mesh adaptation procedure, included in ISIS-CFD, is used, choosing as refinement criterion the flux component Hessian [5]. Thus, for each angle of attack, a specific mesh is obtained. For all simulations, the k-ω SST turbulence model is used.
The major issue in extracting a simple dynamical model from unsteady aerodynamic flows leds to consider “model-free” type of control laws. Two control laws are being investigated: a model-free control [6] [7], and an adaptive control law based on the super-twisting algorithm [8].
The objective of this study is to evaluate the performances of the lift control under a high Reynolds number (Re = 4.7e6), to draw experimental perspectives performed at two different blade scales (i.e. 1/10 chord scale at the LHEEA’s wind tunnel and full chord scale at CSTB’s wind tunnel), and to extend the results obtained experimentally by the authors using the active flow control by micro-jets [9][10].

Controlling wind turbines is generally performed globally (rotor yaw or blade pitch control) to optimize the energy extraction and minimize rotor's loads for rotor's lifetime extension. This means that no information from the blade aerodynamics is up to now taken into account in the control loop while it is well understood that wind inflow interaction with blade aerodynamics can lead to power loss, load fluctuations and noise generation (see e.g. [1] [2]).

This work deals with the development of control algorithms applied at the level of the blade section, considering only local aerodynamic sensing and actuators. The objective is to extract the maximum power from the wind energy by maintaining the aerodynamic lift at its highest value, while limiting load fluctuations using different solutions of control which take into account disturbances from different turbulent inflows. From previous works [3] [4], some control strategies are currently investigated thanks to an experimental bench in the aerodynamic wind tunnel of LHEEA's laboratory (see Figure, [5] [6]). The experimental bench is able to control the lift from micro-jets located at the surface of the blade section. The lift is measured in real-time using ending beam load cell sensors; the control is managed by a STM32 Nucleo board H743ZI2 allowing a 16-bit ADC acquisition.

The difficulty in implementing this control lies in the very uncertain modeling of the lift response, which naturally led to the consideration of “model-free” type control laws. The latter are being investigated: a model-free control [7] [8], an adaptive control law based on the super-twisting algorithm [9] and a robust PID controller [10]. The latter has been implemented considering a rough modeling of the lift response under particular parametric operating conditions of the test-bed. The purpose of this study is to evaluate the performances of each controller under different operating conditions in order to characterize the best operating domain of each control law regarding some criteria like the nominal lift responses, the rejection of high frequency fluctuations, and the robustness towards strong air flow perturbations.

Perspectives include a complementary focus using pitch control in addition to micro-jets actuators in order to cope with a better lift range. Also, more local aerodynamic sensors such as wall pressure sensors or flow separation sensors will be used [11] [12].

Wind energy systems like horizontal-axis wind turbines and vertical-axis wind turbines operate in the turbulent atmospheric boundary layer [1]. Turbulence greatly impacts the efficiency
of these wind energy systems. Thus, it is important to study the turbulent inflow that they encounter [2], [3]. Important statistical quantities that considerably affects the aerodynamic performance of a rotor blade are Turbulent Kinetic Energy (TKE) and length scales in the wind. Experimentally, their effects can be studied in detail by subjecting a Reynolds-scaled
wind turbine rotor or a blade section from a real wind turbine blade to turbulent inflow in a wind tunnel. Different inflow conditions can be set such as homogeneous inflow or gust
inflows [4]. However, experiments are limited in terms of Reynolds numbers (Re), and turbulence intensities (TI). Therefore, properly validated simulations of a
wind tunnel setup known as digital twin, are an important basis for extension of experiments where a large number of test cases can be performed. In wind tunnel experiments, the turbulence properties evolve in downstream of the inflow perturbation. Hence, it is necessary to mimic the same turbulence properties in the simulations to ensure that the test subject is
observing similar turbulent inflows in simulations and experiments. Large Eddy Simulations (LES) [5] are a good option that can be used for performing such simulations as they are capable of resolving energy-containing bigger turbulent structures. However, LES is time-consuming and has a high computational cost. If only statistical moments of the velocity fluctuations are to be considered, then a good alternative to LES are Reynolds Averaged Navier-Stokes (RANS) simulations driven by two-equation eddy viscosity-based turbulence models such as k-omega SST Menter 1994, and 2003 [6], [7]. Finding the correct boundary conditions and closure model is a main challenge here. Torrano et al. (2015) [8] have used two-equation eddy viscosity models (TEM) to simulate the evolution of the TKE in the wake of a grid, however, a comparison with experimental results shows large deviations. In the present study, we reproduced the experimently obtained TKE decay behind a homogeneous grid in RANS simulations. Future works will include other grid configurations as well. Results produced
in this study will be used to produce turbulence perturbations to study control strategies in a simulation environment.

Active flow control devices could significantly improve the aerodynamic performance of a turbine blade and increase its energy production at different flow conditions. Boundary layer suction (BLS) could boost the sectional aerodynamic performance of a blade profile by removing the low-energy flow within the boundary layer. Hence, this type of flow control can result in a net power generation increment [1]. Moreover, a wind turbine blade with BLS could benefit from a delay in flow separation and reduced trailing edge vortex shedding [2,3]. Although the BLS flow control has been extensively assessed for airfoils under low Reynolds numbers [2,4-7], vertical-axis wind turbines [8,9] and small horizontal-axis wind turbines (HAWT) [1-3], the literature lacks works on BLS application to commercial-sized offshore wind turbine blades and wind turbine airfoils with high Reynolds numbers matching the ones at the blade scale. Moreover, how BLS affects the aerodynamic characteristics of the flow near the root section of HAWT blades, where the sectional thickness-to-chord ratio is greater than 30%, and three-dimensional effects are significant, has not been sufficiently investigated.

In this study, three-dimensional CFD simulations have been conducted for the DTU 10-MW RWT blade in a 120°-domain with periodic boundaries. BLS with different suction mass flow rates is applied on the root and tip sections of the blade under various wind speed conditions. SU2 flow solver is used to solve URANS equations along with the k-ω SST turbulence model. Moreover, the two-dimensional airfoil CFD simulation data for the FFA-W3-241 and FFA-W3-480GF airfoils with optimised BLS will be fed into the Blade Element Method (BEM) code OpenFAST. The optimisation study of BLS for these airfoils and its results will be presented in another presentation in detail [10]. The impact of BLS on the net power generation of the wind turbine will be examined with both BEM and 3D CFD results at different flow conditions, and three-dimensional CFD simulation results will be compared with the BEM results. In addition, 2-D airfoil CFD simulation data will be compared with 3-D sectional aerodynamic characteristics at r/R=20% and r/R=80% sections of the blade.

Figure 1 shows how BLS modifies the flow around the root and tip blade sections of the DTU 10-MW RWT blade. It reduces the strength of the trailing-edge wake, increases the suction on the upper surface, and prevents flow separation near the root section.