17:00 - 18:30 Annual energy production - let's get it right!
To improve investor confidence in our energy yield assessments it’s vital that we can demonstrate improvements in the accuracy of our predictions. Only through doing this will be able to access cheaper sources of capital which will in turn reduce the cost of energy.
- Identification of the errors in the model chain.
- How using multiple models can improve prediction accuracy.
- How to use vertical extrapolation models in combination with long term correlation techniques.
- Using the due diligence process to add project value.
Lead Session Chair:
Mike Anderson, RES Ltd.
Claude Abiven (1) F
(1) Natural Power, Strasbourg, France
Printer friendly version: Print
Presenter's biographyBiographies are supplied directly by presenters at EWEA 2015 and are published here unedited
Claude owns masters in Fluid Mechanics and Climate Sciences from Virginia Tech and MIT. He has been working for Natural Power, a wind energy consultancy, for eight years. His work is mainly focused on advanced wind flow modelling for energy yield assessment. His research interests include complex and forested terrain, wake modelling, as well as mesoscale and CFD coupling.
Using multiple models to improve wind speed prediction
With a diverse set of wind flow models available in the industry the quest for the ultimate model for wind flow prediction rages on. More prosaically the question to which model to apply for a given wind farm is often raised.
We propose to examine the effect of combining outputs of several wind flow models on model error and uncertainty, as is often done in climate sciences, but not in the wind industry.
Several model types (coupled CFD, standard CFD, mesoscale and linear) were run on a set of wind farms on which concurrent datasets from 2 masts or more were available.
Wind speed, turbulence intensity, wind shear and wind direction were analyzed.
Errors computed between data and outputs of each one of these models were then compared to errors obtained with a selection of mixes of models (CFD+mesoscale, mesoscale+linear, etc.).
Main body of abstract
It is found that best results are achieved by mixes of models rather than by individual models.
However, it is also shown that mixing models does not necessarily result in optimal values, as some of these models might not be fully independent, sharing similar input data or more fundamental characteristics.
We conclude by discussing the implications of running several models on a single project versus potential added value.
State of modelling uncertainty
High-end wind flow modelling
Ensemble modelling to reduce uncertainty