Publications, Presentations, and White Papers
Submitted to 2021 COLLEGIATE WIND COMPETITION
Final Report from PSU: Winners of CWC's Project Development Challenge using Continuum
Presented by Elizabeth Walls
NREL Wind Workforce Development Webinar Series
As part of NREL's Wind Workforce Development webinar series, an overview of the wind flow model algorithm used in Continuum® is presented. The derivation of this reduced-order Navier-Stokes wind flow model is presented in detail. The theory behind the roughness model used in Continuum® is also explained.
Due to certain simplifying assumptions, the computation time for Continuum is very fast while maintaining a high level of accuracy. The results of a wind flow modeling study show that the RMS of the wind speed error is 1 - 1.5%.
The latest version of Continuum® contains many new modules including time series modeling, MERRA2 data download, MCP, among others. In the webinar, the wind flow modeling module is demonstrated. This includes creating site-calibrated wind flow models, conducting a Round Robin uncertainty analysis, creating wind speed maps, and generating gross turbine estimates.
by Elizabeth Walls
American Wind Energy Association (AWEA)
Wind Resource Seminar, December 2018
A novel approach of assessing a wind farm’s performance using pre-construction and operational met data has been developed and is presented with results of a case study from a wind farm with over 100 turbines. In this technique, a site-calibrated Continuum® wind flow model is created using pre-construction met data. The met data collected during wind farm operation is then imported into the model and the net energy production is calculated at each turbine.
By comparing true energy production to modeled energy production, under-performing turbines can be easily identified and the impact of production boosting technology can be assessed. Additionally, this technique allows for the quantification of the model uncertainty which will improve the confidence level of future project estimates. In the case study, the average modeled energy uncertainty was 4.0%. Finally, by utilizing this method on an annual basis, changes in the wind farm’s performance may be tracked and monitored.
by Elizabeth Walls, August 2015
At the heart of every wind resource assessment is the wind flow model. It describes the wind resource variability across a project area. If the model is flawed or biased, then all subsequent calculations and estimates will inherit those flaws. It is therefore very important for the wind flow model to accurately characterize the wind resource in order for the wind farm’s maximum net energy production to be realized.
Currently, in the wind energy industry, there are a number of commercially-available wind flow models. In this study, three are tested in a side-by-side comparison where each was used to model the wind resource at a project site with eleven met towers. Using the same met data in each software program, wind flow models were generated and then a ‘Round Robin’ (or ‘Leave One Out’) approach was used to test the relative accuracy of each model. The results of this side-by-side comparison are summarized in this white paper.
by Elizabeth Walls, September 2015
In the first version of Continuum, released in January 2015, one of the model assumptions was that the surface roughness was approximately uniform and that only variations in the terrain would have an effect on the wind speed. While this assumption may be valid in some situations, there are many instances when the surface roughness is not constant and should be considered in the wind flow model.
In Continuum 2.0, a surface roughness model has been implemented and is based on the log law shear profile. In this new version, land cover data is imported and is converted to surface roughness and displacement height. To estimate the wind speed across a project area, the variations in the surface roughness and displacement height are analyzed and their influence on the wind speed are estimated.
To test the relative effect of applying the surface roughness model, a validation study was conducted at eleven project sites across the U.S. At each site, the number of met sites varied from four to eleven and, at each site, two Continuum wind flow models were generated: one with the surface roughness model applied and one ignoring the surface roughness. This summary report presents the results of this ‘before and after’ study.
by Elizabeth Walls
Wind Engineering, Volume 39, Issue 3, Spring 2015
ABSTRACT When developing a wind farm, it is very important to accurately define the wind resource distribution across the project area such that an optimized turbine layout can be achieved. To estimate the wind resource distribution, typically, meteorological (met) towers are installed at various strategic locations and the wind speed and direction measured at these sites are used as inputs into wind flow models. Currently, linear and CFD models are most commonly used. Linear models can provide estimates quickly, with little training and at a low cost however, this type of model is well-known to deliver highly inaccurate estimates particularly in complex terrain. CFD models can provide more accurate estimates however they require significant computational time, an expert knowledge level and a much larger financial investment. Also, all commercially-available linear and CFD models are limited to using a single met site in the model creation.
A new wind flow model, Continuum (patent pending), is introduced which is based on a simplified analysis of Navier-Stokes and utilizes data from all of the met sites simultaneously to develop site-calibrated models. The model coefficients, mUW and mDW , describe the sensitivity of the wind speed to changes in the upwind and downwind terrain exposure and are defined for downhill and uphill flow. The coefficients are a function of terrain complexity and, since terrain complexity can change across an area, the estimates are performed in a stepwise fashion where a path of nodes with a gradual change in complexity are found between each pair of sites. Also, coefficients are defined for each wind direction sector and estimates are performed on a sectorwise basis. The site-calibrated models are created by cross- predicting between each pair of met sites and, through a self-learning technique, the model coefficients that yield the minimum met cross-prediction error are found.
A case study is presented where eleven met masts at a complex terrain site were modeled in Continuum. Using the site-calibrated model, the wind speeds were predicted at the met sites and excellent agreement was found between the estimated and actual wind speeds with a correlation coefficient of 0.96 and an RMSE of 0.90%. The largest wind speed estimate error was 1.6% and five of the eleven sites were modeled to within an error of 0.5%. In Continuum, a Round Robin analysis was performed using met subset sizes of 8, 9 and 10 mets where every possible combination of met sites were used to form a model which were then used to predict at the excluded met sites. The RMS error of the Round Robin predictions was ~1.6% for all three subset sizes which confirmed the very good quality, high level of robustness and validity of the Continuum wind flow model.
by Elizabeth Walls and Jack Kline
American Wind Energy Association (AWEA)
Wind Resource Seminar, December 2013
This study, presented at the AWEA Wind Resource Assessment workshop in 2013, focused on analyzing terrain-exposure based models that were developed at 12 different sites across the US and Canada.
Previous studies had shown that there often exists a linear relationship between terrain exposure and wind speed however at least three met sites were required in order to form a meaningful linear regression. In this study, the model coefficients found at the twelve different sites were analyzed and compared and it was found that the coefficients are related to terrain complexity and atmospheric stability. Based on the coefficients found at the 12 different sites, a universal upwind and downwind (UW&DW) model was developed such that wind speed estimates could be formed with a single met site. The model was established based on the twelve sites and, with a single met as the predictor, the wind speed estimate error ranged from 1.7% at sites with simple terrain to approximately 3% at sites with highly complex terrain. Additionally, the model was tested at a site that was not included in the model development and the RMS error of the wind speed estimates was 1.45%.