Multidisciplinary optimisation of wind farms through experiments and data-driven modelling

This project will address two main challenges related to wind power generation. The first challenge concerns the deployment of wind turbines in a wide range of environments due to the turbine size. The extracted wind power, indeed, increases with the size of the turbine, limiting the positioning of large wind farms offshore or in deserted areas.

This project will aim at increasing power generation performance through a numerical optimisation of the turbines' position, thus allowing a reduction in in the wind farm size (at equal extracted energy). Reduced wind farm volumes will be beneficial to mitigate their environmental impact, while benefiting from the versatility to exploit a clean energy source in various locations, e.g., in proximity of urban areas.

The proposed optimisation will also allow us to address the second challenge, which concerns inter-turbine interaction. It is indeed well-known that the wake of an upwind turbine in a wind farm interacts with downwind turbines, lowering the overall wind farm energy yield.

The project will develop a new approach to wind farm optimisation by exploiting turbines with different characteristics (e.g., axis orientation and/or size) to exploit the complexity of the airflow within the wind farm. The PhD student will perform and employ experimental measurements to test and validate the optimisation process, thereby limitations in the flow measurements will arise.

A positioning system will be assembled to allow the automatic movement of a subset of turbine models (while others remain at fixed location) along one or two orthogonal directions, representing a methodological novelty and challenge per se. An in-house data-driven model will also be adapted to be embedded within the optimisation model, to estimate the wind farm performance by exploiting the output of existing analytical models and data from (atmospheric flow) wind-tunnel measurements.

Accordingly, the aim of data-driven model is twofold: (i) supporting measurement limitations by providing an estimate of extracted power (based on given atmospheric flow conditions); and (ii) gaining performance forecasting capabilities of the wind farm. Besides experimental limitations, therefore, the project will include several modelling challenges.

One crucial challenge will be the triple integration between the optimisation framework, the data-driven model, and the input data (from experiments and/or analytical models). Furthermore, the generalisation of the proposed approach to a variety of atmospheric conditions will represent a notable challenge that this project will start facing.