Back to the programme printer.gif Print



Thursday, 13 March 2014
09:00 - 10:30 The model chain: First steps towards tomorrow's technology
Resource Assessment  


Room: Tramuntana
Session description

Recent advances in the software and computational resources available to the wind industry have opened a new frontier; the ‘model chain’. A single approach to such a concept has yet to emerge as the industry standard, although a general idea of a dynamic process across progressively smaller scales is emerging. This session intends to give delegates a broad view of how research institutes and private companies are dealing with this challenge, what the most promising approaches are and which range of applications is foreseen for the coming years.

Learning objectives

  • Understand some of the challenges of multi-scale modelling
  • Get a first glimpse of current approaches in this topic
  • Talk directly to the main players in this research field
Lead Session Chair:
Pep Moreno, Vortex, Spain

Co-chair(s):
Hans E. Joergensen, DTU Wind Energy, Denmark
Michael Brower AWS Truepower, LLC, United States
Co-authors:
Michael Brower (1) F P Philippe Beaucage (1) Jose Vidal (1)
(1) AWS Truepower, LLC, Albany, United States

Printer friendly version: printer.gif Print

Presenter's biography

Biographies are supplied directly by presenters at EWEA 2014 and are published here unedited

Michael Brower is Chief Technical Officer for AWS Truepower, a renewable energy consulting firm. Michael has authored or co-authored several books, including "Wind Resource Assessment: A Practical Guide to Developing a Wind Project" (Wiley 2012). He has also led the development of several new methods supporting the growth of renewable energy, including the application of numerical weather prediction models to resource mapping and forecasting; advanced wind flow and “deep-array” wake models; and grid integration and impact studies. He holds a Ph.D. in Physics from Harvard University, and lives and works in Boston, Massachusetts, USA.

Abstract

A Performance Comparison of Four Numerical Wind Flow Modeling Systems

Introduction

Model chains, consisting usually of coupled numerical weather prediction (NWP) and linear wind flow models, have been used for over a decade for micrositing and resource mapping. They are designed to capture key atmospheric phenomena at relevant scales while keeping computer time manageable. However, only a few, limited studies have assessed whether these modeling systems perform better than the more widely used, stand-alone models such as Jackson-Hunt and RANS-CFD, and under what meteorological and topographic conditions. This research study provides a preliminary answer to these questions, and invites dialog and further research.

Approach

A statistically robust sample of cases is analyzed to assess the relative performance of different wind flow models and modeling systems under a wide range of conditions. This approach is intended less to elucidate details of model performance for the benefit of researchers than to provide guidance for everyday practitioners tasked with choosing the best model for a specific project. Consequently, we run the models the way they are actually used in resource assessment, i.e., over all conditions encountered at each site, and compare mean speed errors, which are the primary driver of errors in energy production estimates.

Main body of abstract

This study compares the performance of 2 stand-alone wind flow models – Jackson-Hunt and RANS CFD – with 2 model chains – NWP-mass consistent and NWP-large-eddy simulation (LES). Modeling errors are evaluated for a sample of 144 tower pairs at 9 diverse project sites, ranging from mountainous, tree-covered terrain to open rolling plains and coastal sites. For each tower pair, each model is adjusted or driven from one tower, and the predicted mean speed is compared with the observed mean speed at the other tower. The same inputs - topography, land cover, wind measurements - are used for each model to the extent possible, and the models are run in their standard configurations. Adjustments to masts are carried out using the standard directional speed-up approach employed in plant design software.

The dependence of errors on inter-tower distance, terrain complexity, tree cover, and other factors is investigated for each model. Based on the sample studied, it is found that the model chains tend to perform better than the stand-alone models by a margin of about 25% overall, and that the difference is most pronounced in complex, tree-covered terrain, and over distances greater than 3-4 km. At short distances, model performance converges, though RANS-CFD and the NWP-based model chains appear to have an edge over Jackson-Hunt at this scale.

Conclusion

Based on this study sample, NWP-based model chains appear to offer significant benefits over stand-alone models in predicting wind resources in complex, tree-covered terrain, particularly at large sites. This is consistent with theory, considering that the time-varying effects of temperature gradients and atmospheric stability, which are fully handled only in NWP models, can become especially important in such conditions. While these findings are informative, other researchers and practitioners are invited to perform similar assessments, so as to enable a dialog that may enrich understanding of the practical application of wind flow models and model chains.


Learning objectives
1) Become familiar with the main types of stand-alone models and model chains and the meteorological phenomena represented in their equations.

2) Gain appreciation of how different meteorological conditions may affect wind flow, e.g., the influence of thermal stability on terrain blocking and channeling.

3) Understand the importance of meta-analyses of many cases for discerning patterns or trends in model performance over a range of site conditions.