Posters

Abstract

In the context of global warming and the coming energy crisis, wind energy should, in the near future, take an important share of the fossil fuels in the world energy mix. In order to increasethe available power of wind energy, the choice of wind energy managers is to gather wind turbinesin wind farms. An issue that arises when wind turbines are close to each other is that they caninterfer aerodynamically. When a turbine is in the wake of another turbine, it collects less energy and undergoes a higher turbulence intensity. These conditions can lead to a substantial waste of wind energy and possibly shorten the life of their components due to higher fatigue loads. A promising solution to compensate these issues is wind farm flow control. Wind farm flow control consists in reducing or redirecting the wake of wind turbines by producing less power or yawing the rotor of upstream turbines, depending on the upstream wind conditions. The recent research around wind farm flow control mainly focus on control strategies in simulation environment. However, it is common knowledge in control systems engineering that a control system without feedback information is not a robust solution within a realistic environment. Therefore, there is a strong interest for the wind farm control community to design efficient estimation strategies. In order to have relevant feedback information for the control strategies to be implemented in commercial wind farms. Most control strategies that can be found in the literature require information on freestream wind conditions. However, in order to have a relevant wind farm flow model that matches the behavior of a commercial wind farm, it is of primal importance to reduce the uncertainties on the properties of the wind farm. A wrong estimation of the wind farm parameters could damage strongly the performance of some wind farm controllers. In this presentation, a dynamic estimator aiming at reducing the uncertainty of wind farm parameters will be presented, in order to alleviate this kind of issue. This estimator is fed with 10 minutes SCADA data and tested in several controlled simulation environments, in order to understand its strength and limits. Eventually, the presented estimation strategy is tested under an extensively studied dataset, from the ANR-SmartEole collaborative project. These results are discussed in the presentation.