Data assimilation for wake-wake interactions

Project Description:

This Research Project is part of the Aura CDT’s Predicting Offshore Wind wake interactions for Energy and the enviRonment (POWER) Cluster.

Large scale wind farms often consist of hundreds of wind turbines with diameters going up to hundreds of metres. The wakes generated by these turbines interact with each other. The accurate modelling of the interaction between the wakes can have significant impact on our ability to optimise the operations of large wind farms and maximise their energy output.

Models of different levels of fidelity are developed in parallel to model wake-wake interactions. Novel semi-analytical wake models provide efficient estimate of the key mean features. High-fidelity simulations such as large eddy simulations (LES) can provide highly resolved three-dimensional turbulence, which are often used to understand the underlying physicsof the flows and to provide detailed databases for the calibration of engineering models.

Some recent research has focused on controlling wind farms for power generation optimisation or power tracking, as wind energy gradually becomes main source of electric power. For example, Bastankhah and Porte-Agel show that yaw angle control can increase power by 17%. However, the control algorithms often introduce additional unsteady modulation to the wakes. For example, it was found by Munters and Meyers that the optimal yaw and induction control strongly oscillates in time. To capture accurately the impact of the unsteadiness is the new challenge for large eddy simulations as well as other modelling approaches, which has not been fully accounted for. For example, Lin, M. and Porte-Agel, F. shows that the usual ADM and ALM models perform poorly in simulations with active yaw control. More generally, as stated in the recent review by Shapiro et.al. on the topic: “Computational approaches that enable higher-fidelity representations under the rapidly changing behaviour of a controlled wind farm remain an ongoing challenge”.

The scientific question behind these new challenges is how to model or parametrise the non-equilibrium features in the wakes (which in this case are introduced or amplified by the controls). This question has long been at the core of wind farm modelling and is one of the main questions being addressed by this cluster. This project intends to focus on a data driven approach, taking advantage of the availability of wind tunnel as well as field data that have been accumulated rapidly. The aim is to synthesise data assimilation (DA) techniques with LES to develop a modelling approach that will improve the understanding and prediction of wake-wake interactions. The application of data assimilation in the context of LES has received only limited research; many questions remain open. The project is composed of several inter-connected objectives.

1. Assessing the deficits of current LES models in the simulations of controlled wind farms.
2. In order to effectively and consistently synthesise data with LES, we will investigate the sensitivity of the wakes with respect to the data. Data enrichment or reduction might be necessary.
3. Developing ensemble Kalman filter (EnKF) based methods to improve subgrid-scale modelling as well as LES prediction for unsteady wake-wake interactions.