The wind farm as a sensor: learning and explaining orographic and plant-induced flow heterogeneities from operational data

The study presents the methodology, implementation, and first results of the experimental test of a lifetime-aware wind farm control (WFC) approach. The novel strategy goes beyond the conventional WFC formulations and aims to maximise profit while ensuring fulfilment of a desired lifetime. Damage due to fatigue is formulated using a cost model and is also limited by a lifetime constraint. The control strategy is based on yaw-induced wake steering and utilises a tool chain of power and damage estimators. The experiment was conducted with three model turbines in a boundary layer wind tunnel, which allowed the simulation of a realistic dynamic variation of the wind direction. The novel strategy is compared to conventional greedy operation, as well as to the classical maximum power strategy. The results show that the novel control strategy can improve the economic performance of the farm.

Section: Flow Modelmentioning

confidence: 99%

First experimental results on lifetime-aware wind farm control

Braunbehrens,

Anand,

Campagnolo

et al. 2024

“…Contrary to method A, wake effects and turbine responses are explicitly modelled. To improve the accuracy of the engineering model, it is also site-specifically adapted with the "wind farm as sensor" method [5].…”

Section: Study Overviewmentioning

confidence: 99%

“…However, using it in its baseline configuration could limit the predictive accuracy of the hybrid approach. Therefore, the baseline model is site adapted following the "wind farm as sensor" approach, described in greater detail in [5]. Historical SCADA data is used to tune some of the model parameters, including the wake velocity deficit and the wake-added turbulence.…”

Section: Engineering Flow Modelmentioning

confidence: 99%

Site-specific production forecast through data-driven and engineering models

Braunbehrens,

Strecker,

Anand

et al. 2024

The present study investigates the predictive capabilities of two site-specific wind power forecasting models. The two approaches make use of a machine learning ensemble model that combines forecasts from several numerical weather predictions (NWP) to match local observations. In the first approach, the training data consists of historical site power production recordings. This yields a black-box model that includes the wind farm behaviour. The second approach is a combination of data-driven and explicit modelling. In the first step, the machine learning ensemble model forecasts the site wind conditions. Then, this input is fed to a site-specific engineering wake model. This hybrid ”grey-box” approach explicitly resolves some farm flow effects. The study shows that the performance of both methods differs over the wind speed range. The purely data-driven model performs better at medium wind speeds, which also occur more often. The hybrid method predictions agree better during periods of higher wind speeds. In the last part, the study showcases the capabilities of the forecasting methods when participating in the energy market. Additionally, the engineering wake model allows for power maximisation through wind farm control.

“…In figure 3 the boundaries of the sections are indicated by the grey dotted lines. Behind the near-wake region (shown as horizontal lines behind the rotor), the far wake expands laterally in each section with a wake growth factor that depends on the local turbulence intensity at the respective lateral side according to equation (12). The streamwise development of the local turbulence intensity at the lateral sides of the wakes is shown in the lower subplot of figure 3.…”

Section: Turbulence Model and Wake Growthmentioning

confidence: 99%

“…Usually either a linear or a squared-root wake superposition is used, although there is no common consensus yet on which combination model yields the best results. In this work the linear wake combination model is employed, as it showed better agreement with field data in previous studies [2,12]. The length and velocity distribution of the near wake was adopted from [8] and depends on the misalignment 𝛾, on the thrust coefficient 𝐶 𝑇 , computed with the average wind speed in the rotor area, and on the turbulence intensity 𝐼 𝑙𝑜𝑐𝑎𝑙 at the rotor, according to:…”

Section: Turbulence Model and Wake Growthmentioning

confidence: 99%

A non-symmetric Gaussian wake model for lateral wake-to-wake interactions

Vad

Tamaro

Bottasso

2023

In this paper, we propose a new non-symmetric Gaussian wake model, which allows for different lateral expansions on the two sides of a wake to account for its interaction with neighbouring wakes. The proposed model is formulated following classical speed-deficit assumptions and momentum conservation. Departing from the existing literature, a non-symmetric Gaussian function is used to represent the velocity deficit in the wake. Accordingly, different wake expansions are assumed on the two sides of the wake, each expressed as a function of the locally prevailing turbulence intensity. The model considers that wake-added turbulence changes with downstream distance; hence, the turbulence intensity on a wake-immersed side of the wake is location dependent. The new model is compared to LES-ALM numerical simulations of three turbines in partial wake overlap. The free parameters of the model describing the wake development are tuned based on the CFD results. Results indicate that the new model provides for a very good agreement of the velocity profiles at different downstream positions, generating an improved representation of merging wakes and their downstream development.