To be effective, wind mapping technology must cover vast areas, typically the size of complete states or even countries. At the same time resolution is important, as topographic features that accelerate the wind to economic levels can be hidden in low resolution mapping.
WindScape is a broad area high resolution wind mapping system that is capable of mapping vast areas at a resolution of 100m or better. Originally developed by Windlab scientists while they were with the Australian scientific organization CSIRO and then extensively developed in-house, the software uses as its starting point inputs from weather data collected around the world.
Atmospheric data – wind speed, wind direction, temperature and humidity – are measured across the globe at 6 hourly intervals. To arrive at a fully 3-dimensional picture of the atmosphere, this data is collected not only at the surface but during balloon ascents that reach heights of up to 20km – in excess of 2000 balloon flights per day around the globe!
This metrological data is then used to construct a picture of the atmosphere for the whole world on a 1/2 x 1/2 degree grid of latitude and longitude. Long-term archives of this data, serve as a starting point for use of sophisticated computer models to determine the wind climate anywhere across the globe, at very high resolution.
Finding the Coonooer Bridge Wind Farm
A Needle in a Haystack
In 2015, Coonooer Bridge Wind Farm (CBWF) was awarded the Clean Energy Council Community Engagement Award. CBWF is a wind energy project located north west of Bendigo, Victoria that is partly owned by landholders neighbouring the project.
This is the first renewable energy project in the country with an ownership structure that includes the local farming community in this way. By all measures the project is an outstanding outcome in terms of what it delivers for Windlab, its shareholders and the community.
Here is how it all started.
How closely do we need to look to find valuable wind farm sites? Atmospheric modelling resolutions of 9km (grid squares that are 9km on a side), as seen here, are fairly common in wind prospecting.
Modelling resolutions similar to this can give what looks like a fairly detailed picture of the wind climate when viewed from a perspective that takes in very large areas such as whole states as in seen here.
However, to be useful in wind prospecting we need much more detail.
If we enlarge the area enclosed in the rectangle in the previous image we can see that, though the detail has increased, the same holds even for 3km resolution map, with the map really only providing information that is what we know anecdotally – it’s more windy over higher terrain. We need more detail.
At 1km resolution, if we once again enlarge the area enclosed by the inner rectangle in the 3km resolution image above, we are starting to see where a needle might be found, but it’s still really anybody’s guess. Remember that we need to be precise enough to identity individual properties, hills, and even potential wind turbine locations.
Up to this point, WindScape has used atmospheric models that work in a way similar to a weather forecast model, to narrow down the search. However, the processes that make air accelerate over small hills (say a few km in size) are quite different, working more like air flow over an aeroplane wing than a weather system.
The outer rectangle in the figure above is roughly 20km on a side, while the inner rectangle is 10km on a side and is the same rectangle in the figure below.
This is where the WindScape technology changes gear and really hits its stride. Here WindScape uses what is called a microscale model to calculate how the wind speeds up over small topographic features that are well hidden in 3km and even 1km resolution calculations.
This proprietary model uses complex mathematical principles, not entirely dissimilar to those used to design an aeroplane wing, to make the calculation extremely efficient. Using this model enables vast areas to be searched at very high resolution for potential wind farm sites.
At 100m resolution we have found what may be the needle in the haystack that is all 200,000 square kilometres of Victoria! But wind farm development is more than finding the best wind resource.
So far all that we’ve examined are computer models of wind speeds – but what does this landscape actually look like? For example, is the area dotted with quarries, is it heavily vegetated or is there a town of people right in the middle of our proposed wind farm site?
This aerial photo, of the same area covered by the WindScape image above, shows that the landscape is actually one of extensive farming.
But who owns the land, how close are residences from potential turbine locations and what other environmental, planning and cultural issues need to be considered? These are critically important to the development process. In the image to the left, the pink hatched areas show 2 kilometre exclusion zones around residences, dark green areas are nature or scenic reserves with the light green area being a vegetation exclusion. Finally, in the remaining area, the white dots are turbine locations with the connecting access roads shown in blue.
The ideal outcome of this desktop work is a constructed project, in this case Coonooer Bridge Wind Farm. Following consultation with local communities and landholders, on-site monitoring of actual wind speeds, completion of environmental impact and other development stages, the construction of roads and foundations can go ahead, followed by turbine erection.
Outstanding projects like CBWF start with a broad area, high resolution view of the wind resource. This maximises the number of options in the development process, which can be quite complex. By using sophisticated software, developed in-house, together with the right expertise, advanced mapping and remote sensing technologies, Windlab can find the best locations for wind farms anywhere in the world – needles in haystacks!
Wind resource modelling has its origins in the 1970’s. Since then much has changed in terms of the models used and the methods for wind data collection. However, one thing that has not changed significantly is the overall methodology used to create a high resolution wind map across a wind farm for the purpose of optimising turbine locations.
Windlab has been working to develop a much more modern and accurate methodology that makes optimum use of all on-site measurements combined with state-of-the-art regional-scale and very high resolution flow modelling. This modelling system is called WindScape Hybrid Deterministic Statistical Method (HDSM).
The Problem with Industry Standard Wind Resource Assessment
Two views can be taken of modern wind resource assessment – top-down and bottom-up. In the former a regional scale atmospheric model is used to calculate the broad-scale wind patterns. These calculations are typically driven by global weather data that is collected and stored across the globe. Calculations of this type can accurately represent broad scale weather patterns that occur over several hours to several days.
The bottom-up view is much more like the traditional method used in wind engineering. Wind measurements are collected at one or a few locations across a prospective wind farm and the wind climate is extrapolated to the locations of proposed wind turbines. Fine scale flow models are typically used in both of these methods to account for fine scale topographic features and surface types that affect the wind climate locally.
The advantage of the top-down method is that it can account for broad variability in the wind climate, which can be significant across distances of several kilometres – as many wind farms span. Its disadvantage is that the calculation, though driven by global meteorological data, does not contain information from locally collected measurements. Conversely, the advantage of the bottom-up or conventional wind engineering methodology is that it is the direct product of local measurements. This is clearly important in technical and financial due diligence as it encompasses both high quality local measurements and wind climate. However, what this method lacks is the ability to account for the broad scale variability in wind climate noted in the description of the top-down method.
WindScape HDSM is a new technology that combines meso- and microscale models along with measurements. WindScape HDSM calculates an optimised wind resource estimate that accounts for broad scale variability of the wind climate across a wind farm site as well as more commonly calculated microscale variations due to topographic forcing, surface roughness changes and distributed drag within vegetative canopies. WindScape HDSM uses measurements and statistical transfer functions to bridge the gap between mesoscale and microscale model calculations and in doing so takes advantage of the best aspects of currently available modelling technology and valuable on-site measurements.
WindScape HDSM Features
Exact in wind speed or energy at measurement locations
Solution varies smoothly and in a physically consistent way between measurement locations
Contains the effects of stability down to 300m resolution
Maintains full probability distribution of wind resource at each grid point (no Weibulls)
Yields mast-to-mast and all-but-one-mast energy yield errors as part of calculation
Includes mesoscale variation in wind climate across large and complex sites
Can include simple linear or CFD calculations of speedups
Can be refined as better mesoscale and/or CFD calculation become available
The Science and Technology Behind Wind Farm Prospecting and Development
Good technology comes from sound science. Windlab scientists are in the business of creating good technology from sound science and using it in ways that help wind farm development by increasing accuracy and reducing risk associated with the development process. Technology development is a long-term investment. Windlab scientists have been doing it for more than two decades and are still pushing forward with developments in areas like computational fluid dynamics, turbulence modelling and data analytics using machine learning.
The following outlines a few areas of current research and development.
Computational Fluid Dynamics (CFD)
In the early days of wind farm development, the available flow modelling tools were very modest compared to what the industry has today. The reason for this was a lack of computing power to carry out calculations in a commercial context in a timely and accurate fashion.
Nearly two decades ago, the scientists that became the founders of Windlab were working on this problem, developing new computational flow models for use in wind farm development. These models removed the limitations inherent in the then available models as computation became vastly faster and cheaper. At the same time these models became more accurate by including more and finer details about the factors that affect the wind flow across a wind farm.
Today at Windlab, a typical calculation used in wind resource assessment is done on a high performance computing cluster using a high speed interconnect to distribute calculations across hundreds or even thousands of CPU’s. This allows Windlab to solve larger, more difficult problems in fluid flow.
Many areas of science and industry are currently making significant advances through machine learning. These applications run from home automation systems, driver-assist in many new cars or even self driving cars, natural language translation, recommendation systems, face recognition and the list goes on. Organisations that hold large data assets have a significant advantage. Windlab maintains in excess of 70 met towers and solar measurements. Windlab’s asset management activities also provide Supervisory Control and Data Acquisition (SCADA) data from two wind farms. This will soon be increasing with the addition of the Kiata wind farm and the Kennedy Energy Park combined Wind, Solar and Storage (WSS) project.
Diagnosing Turbulence from Remotely Sensed Data Using Neural Networks
Atmospheric turbulence, the variation of the three dimensional wind speed over short periods – from a few seconds to a few minutes – can cause significant wear and tear on or even damage to wind turbines. Because of the localised nature of turbulence, to measure atmospheric turbulence across a wind farm site, one would need to measure in several locations down wind from potential sources of turbulence, for example steep topographic slopes facing backwards in the wind flow. In this context, in addition to their significant expense, the fixed location of an anchored meteorological tower is also a disadvantage. To get around this problem remote sensing methods like SODAR ( SOnic Detection And Ranging) are used. However, SODAR’s are less capable of accurately measuring atmospheric turbulence than traditional fixed tower measurement methods. But if we locate the SODAR next to the meteorological tower for a period of time sufficient for machine learning algorithms to learn the relationship between the fixed tower measurement and the SODAR measurements of the same turbulence, the SODAR measurements can be used with confidence. With this learned relationship in place, the SODAR can then be moved to new locations across the wind farm to understand where turbulence is a risk and avoid those areas.