From wind to loads: wind turbine site-specific load estimation with surrogate models trained on high-fidelity load databases

Dimitrov, Nikolay Krasimirov; Kelly, Mark; Vignaroli, Andrea; Berg, Jacob

doi:10.5194/wes-3-767-2018

Cited by 96 publications

(96 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Calibrating this function against higher fidelity simulations allows replacing the computationally expensive models with a quick and useful approximation. In Dimitrov et al, several wind turbine load surrogate model choices are compared, and the polynomial chaos expansion (PCE) is recommended as the potentially best choice out of the models considered. Further, in Schröder et al, ANNs are shown to match or exceed the performance of the PCE on a similar problem.…”

Section: Formulationmentioning

confidence: 98%

“…For additional details regarding the calibration and use of PCE models, please refer to Sudret and also Murcia et al and Dimitrov et al for an example application to other wind energy‐related problems.…”

Section: Formulationmentioning

confidence: 99%

“…The training set should also cover the variable space to the best extent possible. For the present problem, a stochastic collocation approach is adopted similarly to Dimitrov et al and Murcia et al The sample space is represented in terms of independent, uniformly distributed variables in the range [0,1], which represent the cumulative distribution functions of the physical variables. The mapping from physical to uniform space is done by means of a Rosenblatt transformation .…”

Section: Formulationmentioning

confidence: 99%

“…Conditional independence is assumed for all other variables (eg, it is assumed that the water depth is not dependent on wind shear). The conditional distributions and variable ranges for wind speed, turbulence, and wind shear used in the present study are based on Dimitrov et al A suggested joint distribution model for wind speed, significant wave height, and wave peak period is the one provided by Johannessen et al An example with bounds for all presently considered variables is given in Table . Having the joint distribution and variable bounds in place, a training sample can be generated using a pseudo‐Monte Carlo (pseudo‐MC) approach based on a low‐discrepancy sequence (eg, Halton or Sobol sequence).…”

Section: Formulationmentioning

confidence: 99%

“…When deployed, the surrogate model can provide almost instantaneous load and power maps for various wind farm layouts and wind conditions as long as the same wind turbine type is used. The process of building and training the surrogate model is similar to the one defined in Dimitrov et al and is outlined below in terms of the necessary steps: Define the variable space: which variables are to be included in the training, what are their ranges, and if there are any dependencies between variables (eg, the turbulence range may depend on the wind speed). Generate a random sample covering the input variable space in a uniform way (see Section ). Carry out load simulations for each point in the sample set defined in point 2. In order to account for seed‐to‐seed uncertainty (Section ), carry out multiple simulations at each sample point and collect the statistics (mean and standard deviation) of the output variables for each point. Train a surrogate model that based on the inputs defined in point 2 can predict the outcomes of the simulations from point 3. …”

Section: Formulationmentioning

confidence: 99%

See 4 more Smart Citations

Surrogate models for parameterized representation of wake‐induced loads in wind farms

Dimitrov

2019

Wind Energy

Self Cite

View full text Add to dashboard Cite

A procedure for mapping wake-induced load predictions computed with the dynamic wake meandering model to a computationally efficient surrogate model approximation is defined and demonstrated. Using the mapping function, the load variation can be efficiently estimated for a wind farm with arbitrary layout. The resulting load assessment procedure provides continuous, differentiable output with known analytical derivatives and can be used for applications such as wind turbine layout optimization, estimation of turbine lifetime, and uncertainty analysis. KEYWORDS layout optimization, loads, neural networks, polynomial chaos, probabilistic, surrogate, wake, wind farm 1 | INTRODUCTIONThe wake-induced load effects experienced by wind turbines within a wind farm are an important design factor, which can impose restrictions on the wind farm layout or the turbine design choices. Therefore, wake-induced load analysis is a central part of wind farm planning and design. One of the pioneering approaches was the so-called Sten Frandsen model, 1 which takes wake effects into account by assigning an equivalent ambient turbulence level, which should cause the same fatigue load accumulation as the actual wake-induced turbulence and deficit.The Sten Frandsen approach is widely popular due to its simplicity, but it typically results in conservative load estimates. 2 A more advanced method is the dynamic wake meandering (DWM) model, 3 which provides an engineering approach to including a wake deficit in the turbulence box used for aeroelastic load simulations. The capability of the DWM model to predict wake-induced load effects has been validated for specific turbines in a wind farm, 2,4-6 as well as on turbine prototype test sites. 7 In the latter, the DWM model was shown to have superior performance than other engineering models. A table-lookup-based approach for mapping wind farm power output based on the DWM model has been demonstrated by Keck and Undheim, 8 while a procedure for mapping the load variation within a wind farm is shown in Galinos et al. 9 The aim of the present study is to define and demonstrate a simple procedure that maps the outputs of the DWM model in a computationally efficient surrogate model approximation so that the load variation can be efficiently estimated for a wind farm with arbitrary layout. The resulting quick load assessment provides continuous, differentiable output and can be used for applications such as wind turbine layout optimization, estimation of turbine lifetime, and structural reliability analysis. The main hypothesis is that the wake-induced effects on loads can be described by means of variables related to the wind farm geometry and the ambient conditions and that a surrogate model calibrated by machine learning from a large load simulation database can be used to efficiently quantify such a relation. The parameterization of wakeinduced load effects is obtained under the assumptions underlying the DWM model, ie, that the wake deficit behaves as a passive tracer and follows the transver...

show abstract

Section: Formulationmentioning

confidence: 98%

Section: Formulationmentioning

confidence: 99%

Section: Formulationmentioning

confidence: 99%

Section: Formulationmentioning

confidence: 99%

Section: Formulationmentioning

confidence: 99%

See 3 more Smart Citations

Surrogate models for parameterized representation of wake‐induced loads in wind farms

Dimitrov

2019

Wind Energy

Self Cite

View full text Add to dashboard Cite

show abstract

Analysis of the influence of climate change on the fatigue lifetime of offshore wind turbines using imprecise probabilities

Hübler

Rolfes

2020

Wind Energy

View full text Add to dashboard Cite

When discussing the connection of wind energy and climate change, normally, the potential of wind energy to reduce green house gas emissions is emphasised. Hence, effects of wind energy on climate change are analysed. However, what about the other direction? What is the impact of climate change on wind energy? Recently, the effect of a reversal in global terrestrial stilling,that is, an increase in global wind speeds in the last decade, on the wind energy production has been analysed. Certainly, knowledge about potential changes in energy production is essential to plan future energy supply. Nonetheless, at least similarly important is the effect on loads acting on wind turbines. Increasing loads due to higher wind speeds might reduce wind turbine lifetimes and yield higher costs. Moreover, especially for already existing turbines, it might even affect the structural reliability. Since the impact of climate change on wind turbine loads is largely unknown, it is studied in this work in more detail. For this purpose, different existing models for predicted changes in wind speed and air temperature and their uncertainties are used to forecast the environmental conditions an exemplary offshore wind turbine is exposed to. Subsequently, for this turbine, the lifetime fatigue damages are calculated for different prediction models. It is shown that the expected changes in lifetime fatigue damages are present but relatively small compared to other uncertainties in the fatigue damage calculation.

show abstract

Data‐driven modeling for fatigue loads of large‐scale wind turbines under active power regulation

et al. 2020

View full text Add to dashboard Cite

Advanced control methods for coordinatively optimizing active power dispatch and fatigue load distribution of wind turbines (WTs) in mountain wind farms, require the fatigue load modeling that has not been fully studied. This study proposes a data‐driven modeling method for fatigue loads of large‐scale WT under active power regulation. Firstly, load simulations based on Monte Carlo approach are conducted to overcome the uncertainty of fatigue load caused by turbulent wind. Subsequently, the target components of WT are selected according to the fatigue load dataset distribution at various power setpoints and the sensitivity to active power. Specifically, the Jarque–Bera statistic is used to evaluate the distribution characteristics of datasets; the single‐value processing determines the valid value through the maximum probability of the Gaussian kernel density distribution of the dataset; the Morris screening is employed to select the valid value with high sensitivity to active power regulation. Afterwards, arbitrary polynomial chaos expansion and support vector regression are performed to establish the data model of fatigue loads, respectively. Finally, the proposed method is applied into a 1.5‐MW commercial WT model designed by the manufacturer through the Bladed software. The obtained data‐driven model establishes a basis of simultaneously optimizing active power dispatch and fatigue load distribution, making it possible to reduce energy cost in mountain wind farms.

show abstract

From wind to loads: wind turbine site-specific load estimation with surrogate models trained on high-fidelity load databases

Cited by 96 publications

References 42 publications

Surrogate models for parameterized representation of wake‐induced loads in wind farms

Surrogate models for parameterized representation of wake‐induced loads in wind farms

Analysis of the influence of climate change on the fatigue lifetime of offshore wind turbines using imprecise probabilities

Data‐driven modeling for fatigue loads of large‐scale wind turbines under active power regulation

Contact Info

Product

Resources

About