Gemma Church finds out how simulation software aids and optimises the decisions behind siting a wind farm
Wind turbines are popping up all over the planet. From massive offshore wind farms to single turbines, the choice of where to put these structures, and how to optimise their power output, is based on a vast range of variables.
The first requirement is to find a site with a suitable wind profile. The site’s suitability must also be assessed in terms of its vicinity to the power grid and geography. Then, finally, the wind farm must be optimised to maximise the power output and minimise expenditure. Simulation software is perfectly suited to aid engineers as they evaluate, design, and build wind farms within tight budgetary, environmental, and feasibility parameters.
Wind forecasting is a key consideration. The terrain, wind directions, weather, and a wide range of additional factors must all be taken into account to make wind profile projections.
A common technique when siting wind turbines is to place them at the summit of a steep hill or close to ridges that overlook the surrounding landscape. However, complex terrain modelling is often required to eliminate the risk of wind flow separation, which will produce excessive turbulence. Such turbulence could damage the turbine, so it is set to shut down automatically if the conditions are considered too hazardous.
Rough ground, or unusual local conditions, can also result in turbulence. This disturbs the airflow, preventing efficient operation and causing wear and tear on the turbine.
One method of gauging site suitability and forecasting wind is to construct a met mast. This temporary structure is set up at the potential site and gathers data on the conditions, including strength and direction of wind, over an 18-month to two-year period. This data is fed into the simulation to predict whether a site is suitable in terms of how much power the proposed farm is capable of producing.
Met masts collect a huge amount of data and, as the survey sites can be quite large, multiple masts or alternative equipment, such as anemometers, are sometimes required. Graham Dudgeon, principal industry manager for utilities and energy at MathWorks, said: ‘The customers have to ensure they are putting in the right number of devices to determine how that site might perform if a wind farm was there. It is very much a big data and data analytics problem.’
Simulation software uses one of two general methods to analyse such data and determine a site’s suitability. The first relies on general approximations, to provide a quick analysis of the total power the wind farm can produce.
Such simple simulations are quicker and cheaper to produce, compared with more in-depth analysis, but this could be a false investment for companies looking to construct large, costly wind farms.
It all comes down to the level of accuracy the customer requires, as Jim Ryan, energy and power industry director at CD-adapco, explained: ‘It is always a judgment call for engineers and their companies to decide how much fidelity you need in your simulation projections. If a 10 or 20 per cent error is acceptable – and it may be if you just want a ballpark idea – then that’s acceptable. But if you want to go beyond that, then you need to consider investing more time and money in the simulation software and the hardware involved.’
CD-adapco’s STAR-CCM+ software was used recently to design the layout of a 43-turbine wind farm in Northern Germany, to maximise the annual energy produced (AEP) by the farm. The software compared the wind flow and power production for multiple layouts, including different wind directions, and the effects of terrain and multi-turbine wakes.
The model optimised the final wind farm design, with the modified design improving the AEP by 8.5 per cent, compared with the original, proposed design. The simulation has since proved its worth with a closely matched projected – versus measured – AEP, representing an accuracy in the simulation of between two and four per cent.
Such high accuracy simulations are a necessity when designing a structure that needs to be right first time. Ryan said: ‘You cannot move a wind turbine once it’s in the ground. You need to simulate every angle the wind could possibly come from and what effects such changes have on the power production. You also want to know what happens if there are doldrums where the wind goes away and, on the flip side, if the wind goes beyond a threshold considered dangerous.’
‘Wind simulation is about studying different scenarios, not just one, and doing it fast, reliably and repeatedly,’ he added.
The data from a wind forecasting survey alone does not tell engineers what the wind profile will look like in the presence of a wind farm, so complex aerodynamics and CFD models are also needed to complete the forecast.
These simulations must take on board a wide range of variables to be assessed by a wide range of specific users – from aerodynamic engineers profiling the wind, to mechanical engineers designing the mechanical aspects of the turbines and their blades and the electrical engineers who are concerned with the power electronics and the connection of the turbines within the farm.
Communication between different engineering disciplines is vital, as Dudgeon added: ‘One of the fundamental ways to reduce the risk of developing a wind farm site is making sure those interdisciplinary teams can communicate their engineering information very effectively.’
Vestas Wind Systems has taken a different approach from the conventional met mast method, instead using historical data from met offices around the world to create a wind simulation of the entire planet. The data is made up of 35,000 measurements and creates a virtual atmosphere, which can model wind environments down to a square kilometre resolution.
When evaluating the best wind farm location, which offers cost savings – in terms of manpower and the mast expenditure – historical data is based on 15 years of data, reducing the likelihood of anomalous results due to freak weather patterns that may occur in the 18 month stint for a met mast, according to Vestas.
Anders Rhod Gregersen, chief specialist at Vestas Wind Systems, said: ‘We can tell what the weather is like in your back garden, for example, using this data without any physical measurements ever being done there. In real terms, you can now evaluate any site around the world in a few minutes and at a fraction of the cost.’
Gregersen added: ‘It is very different from doing a simple extrapolation – this modelling technique involves complex weather modelling based on thousands of differential equations.’
The complexity of any wind farm simulation, including those based on historical data and the met mast method, continues to grow as turbines from around the world gather more data, and this new information needs to be coupled to the existing data to improve the accuracies of the simulations further. Gregersen said: ‘Previously, it was like we were driving a fast car on the motorway but we were steering by looking through the rear-view mirror. Now we are looking out at the road ahead and steering not just for what we see, but what is over the horizon.’
This is achieved by harnessing the simulation power of a HPC environment to manage the ‘tsunami of data coming at us’, according to Gregersen, who sees data handling capabilities as the biggest challenge facing the wind energy simulation space. Gregersen added: ‘The discussion is becoming fact-based so we know what will happen when we set up the wind turbines, before we actually do it.’
Noise is a primary concern for those living near a proposed site. Detailed simulations are required to reduce this impact, and to keep wind farms operating within reasonable noise and vibration levels. While noise is more of a concern to the general public, vibration effects can, for example, adversely affect sensitive scientific equipment.
Xi Engineering Consultants specialises in complex noise and vibration issues. Dr Brett Marmo, technical director at Xi Engineering, said: ‘The key challenge we have when modelling the acoustic part of the model is the mesh size. The size of mesh elements is dependent on the wavelength of the sound being modelled – as the frequency of the sound of interest increases, the wavelength and mesh element size reduces.’
‘At high frequencies the number of mesh elements required to model the air surrounding the turbine becomes very large (tens of millions) making the computation very difficult. Furthermore, as the size of the turbine increases the mesh size also increases,’ Marmo added.
Xi Engineering recently simulated the grinding of teeth within the gearbox of a wind turbine, as this component can cause vibrations, that are perceived as excessive noise. Studying the design data, Xi’s engineers found the gearbox was the source of the noise and vibration, and that it was being amplified by the tower’s steel skin.
Using a Comsol Multiphysics model of the structure, they were able to locate hot spots where the noise was being amplified and test various solutions. The solution was to use a specific material to coat the inside of the tower and reduce the amplification. Due to the material’s high cost, the simulations determined the least amount of material necessary to keep the noise levels below the required standards.
An obvious way to eradicate noise concerns is to site wind farms in isolated locations, such as offshore positions. Fluid flow models become even more important for an offshore environment, as CD-adapco’s Jim Ryan explained: ‘We see even greater value proposition for the offshore wind farm developers because of the way our software works. Our software simulates fluid flow, so an onshore wind turbine is only affected by the wind but an offshore site is affected by two fluids – the wind and the water.’
The effect of the ocean adds a further layer of complexity. Simulations must also take into account the mooring, balance, dynamics, and servicing of the turbines in the aqueous environment, and how this will affect the site’s power output and productivity. Ryan added: ‘The risk is higher because, if the turbines fail to perform as expected, replacing them or servicing them is frightfully expensive – far more expensive than doing it onshore. The need to do it right the first time is compounded.’
The pressure to provide a cost-effective offshore wind farm often comes down to the water depth. Shallower sites are the obvious choice, from the pricing perspective, as the waves are not as severe and cause less damage to the platforms; the site is also more stable as the turbines are based on the seafloor, instead of relying on more novel stabilising techniques; and there are fewer power transmission concerns as the site is closer to the shore. But offshore farms are often forced to move further away from the coastline because of their effects on local communities.
This puts further pressure on the simulation software, particularly the transmission effects. The offshore wind farms are often connected through high voltage direct current (HVDC) systems. Dudgeon added: ‘[Customers] have to extend the simulation of the wind farm beyond just the wind farm. They have to model the transmission technology and part of the electric grid.’
Getting to the grid
Transferring the power generated by a wind farm from the site to the grid is a challenge for all wind farms – both on and off the shore.
A novel piece of research using data analysis and graphing software could resolve this issue. High population regions, which also have a high demand for energy but are close to the grid, may be suitable for wind energy production, despite having a low average wind speed, the research revealed.
A three-dimensional animation from OriginLab was used graphically to represent complex wind-speed dynamics for South Palm Beach in Florida, USA. The simulations found the investigated location had enough predictable wind supply to contribute to meeting the area’s peak demands.
The researchers needed a unique data-display technique, as their nonlinear dynamic analysis was an untested approach to wind power evaluation. This differed from conventional approaches to analysing wind-speed data, as Ray Huffaker, professor of Agricultural and Biological Engineering at the Institute of Food and Agricultural Sciences at the University of Florida, explained: ‘Conventional probabilistic wind-project evaluation approaches remove patterns from wind-speed data, and consequently are not equipped to detect wind-power patterns empirically. Our work met a critical need in project evaluation to begin modelling wind-speed dynamics that accurately reflect real-world patterns.’
The researchers investigated the commercial viability of Sugarland Wind, which is the first utility-scale wind project proposed in Florida. The proposal was for a 200 megawatt wind farm with 114 turbines spread across 12,887 acres of sugarcane fields in the Everglades Agricultural Area. Huffaker said: ‘The project received strong community support for contributions to clean energy in the region, community development and the agricultural economy, since sugar cane farmers would lease space in their fields for turbines.’
‘We found empirical evidence that wind-power patterns at the project site match up well with regular daily and seasonal electricity demand patterns,’ he added.
Once the site has been agreed, the exact details of the wind farm must be optimised in terms of the turbine model, type and configuration.
For a wind farm, you are moving away from the small confined areas CFD simulations usually deal with to a grand scale simulation, covering kilometres of open terrain. Turbulence is, again, a key consideration, alongside wake effects between turbines. Basic turbine wake models depend on the turbulence level and thrust of the turbine. Thrust is the downwind hitting the turbine, which directly converts to a loss of wind speed in the wake. The wake from one turbine is detrimental to the turbulence of the other turbines and to wind speed in the farm. The separation and placement of the turbines must be carefully modelled using large eddy simulations and full incompressible flow models.
It is not just a case of choosing the equipment you do need, but the equipment you do not need, as Dudgeon said: ‘If you have a great deal of confidence in your simulation model, you can get to a much more optimised architecture where equipment that you think may be needed – to match a production goal or a grid requirement – may not actually be needed.’
‘The technology choice has a significant bearing on how you put a wind farm together, especially on how you manage a wind farm and operate each turbine to make sure your wind farm, as a whole, is doing what you expect,’ he added.
From turbine selection, to electromechanical modelling or evaluating unusual sites, simulation software continues to power the decisions behind where wind farms will be located within the stringent rules and regulations governing these structures.
Those looking to build wind farms have to solicit investment, win funding and deliver on their promises. It is a high risk scenario and the pain of making mistakes is real and recurring. Solid simulations mitigate this risk when choosing where to put a wind farm.