Dr. Frank Neumann from the University of Adelaide shares his views on wind turbines

Beyond Zero's Matthew Wright talks to Dr. Frank Neumann, from the University of Adelaide, about how the optimal placement of wind turbines in a wind farm increases it's total output.

Beyond Zero intervew Frank Newman

download

Transcript

Matthew: For today’s interview we’re lucky to have Dr Frank Neumann from the University of Adelaide. He is an expert in using evolutionary science to work out the placement of wind turbines. Basically, a wind farm is a group of wind turbines that are erected in close proximity to each other. The idea is to generate electricity and they do that very well. But how can you actually get more electricity out of them? Well that’s a process of choosing the optimal distance between the turbines. We are going to discuss that today.

Hello Frank. It’s great to have you. Thanks for joining us from Adelaide. I understand you are previously from the Max Planck Institute in Germany. How long have you been here in Australia?

Frank: Basically just for six months. I started in January at the University of Adelaide. It’s a great place to be.

Matthew: It’s funny – we talk about the brain drain of our leading researchers from Australia to other countries – but it seems like people from Germany all want to come and work in Australia.

Frank: Yes, we have a very good optimisation group here in Adelaide. That’s basically what attracts me. It’s a very good environment to work on optimisation techniques and also in the field of renewable energy.

Matthew: Now, in terms of renewable energy, you’re looking at using computational algorithms to place wind turbines. Can you tell us a bit of the history of how you got involved in renewable energy. What did you originally do and when did you become involved in renewables?

Frank: I have worked on optimisation algorithms for something like ten years, in particular, evolutionary algorithms. I would basically take solutions [to specified problems] and try to improve them over time [using evolutionary algorithms, processes that mimic evolution in biological populations]. Last year I was visiting some colleagues at MIT in Cambridge, USA. We talked a bit about wind turbines and wind optimisation. Then we realised that there are not many computer science techniques, in particular optimisation techniques, used in this field. We thought that we could bring our expertise into this field and use our techniques to improve placement of wind turbines.

Matthew: So you have only just started applying your skills in computational algorithms to wind farms in the last couple of years. Is that what you are saying?

Frank: Yes.

Matthew: Did you have a personal interest in renewable energy prior to that? Were you always a supporter of renewable energy or was it something that just came along?

Frank: Well, I have been interested in renewable energy and climate change for the past ten or twenty years. But there was never a real opportunity to bring my knowledge into this field. Last year I thought it would be really great if I could just grab my techniques, and the methods that we use in our field, and carry them over to make some impact on renewable energy. I think it’s very important and, as you know, a very hot topic nowadays.

Matthew: When you met the people at MIT did you first have to familiarise yourself with how wind turbines were being conventionally laid out and what was happening in existing farms around the world?

Frank: Yes, that was one thing, but we started off with some very basic things – how wind turbines basically work, how they are modelled, what wake effects are, and so on. We went through all the relevant literature, basically from engineering – when you want to bring in a different discipline it’s a bit harder than staying within your own one – but, when we looked at the research that has been done, we had the feeling we could improve the state of the art, basically by throwing in our algorithms. We also thought it would be possible to have wind farms with a larger number of turbines than are currently used.

Matthew: Can you explain the method for rolling out conventional wind turbines and how you place them in the field?

Frank: It always depends on the criteria that you have. If it’s terrain then you might look at height and where you want to place the turbines. You might look at cable length. People look into different areas. We had the feeling that there are a lot of criteria but no large-scale optimisation.

Matthew: I understand there are two potential outcomes that are wanted. One is where there is a constrained amount of land available, which would be more like a German or European scenario. An example is that the German company ENERCON is offering a service where they actually try to put more wind turbines on the same site. Each wind turbine produces less energy but overall the wind farm produces more energy. The alternative is somewhere like Australia, which is not space-constrained, where you don’t want any wake effects or any interference from the turbines and you want to maximise the output per turbine. Which are you looking at?

Frank: Our setting would mostly be related to something you would see offshore, where you have some space and you want to place the turbines and look at wake effects. This would relate to the European scenario, where the question is how can I get greater benefit from my wind farm by placing additional turbines and how do I place them. This setting also relates to the situation in Australia where I don’t think it’s necessary to have the turbines so far apart that you don’t have any wake effects. If the wake effect problem isn’t that big, then you don’t need to have the turbines as far apart as they could be.

Matthew: I understand the standard rule of thumb in the industry at the moment is a distance between turbines of 5D, five diameters of the wind turbine blades. How does that rule of thumb stack up versus some of the initial results you’ve got in your work?

Frank: We want a safety margin in our simulations. First, we look at the area and choose a distance between turbines of 5D or something like that. Then we simulate the wind farm and calculate the wake effects. Then we move the turbines around, such that the wake effect is minimised. Depending on the wind settings and on how the turbines are placed in relation to each other, you either get larger or smaller wake effects.

Matthew: Can you tell us what wake effects are, in case listeners aren’t across wind technology. What are wake effects?

Frank: Suppose you have wind coming in to one turbine and behind that turbine is another turbine. This second turbine can’t get as much wind as the first turbine and therefore can’t generate as much energy. If you have a wind farm, then, over the course of a year, wind comes from all directions. You might have some model of how the wind is distributed on the site and you want to minimise the effect where one turbine ‘shadows’ the wind for other turbines.

Matthew: Obviously in an array or a set of wind turbines in the sea, you’d tend to go more for a grid approach. But I understand the placement of turbines in some wind farms depends on where the prevailing winds are coming from. The majority of the wind might come from the west or alternatively from the south. So turbines are placed five diameters apart in some directions but not in others. That is, they tend to have different spacing depending on which direction the wind is coming from.

Frank: That’s basically what we would put in our model. We would have a model of the wind direction and then we would calculate the wake effect for one particular setting. Then we would change the placement of the turbines and see whether we can get something that is better. Even if the wind only comes from two or three directions, it’s not clear what the optimal placement of turbines would be because some of these wake effects are very complicated. If you want to calculate these effects, then just looking at the wind directions is not enough to determine the optimal placement of the turbines. It’s really hard to understand that.

Matthew: We’re speaking to Dr Frank Neumann from Adelaide University who is an expert in computational algorithms and he’s actually applying that to wind farm placements, the placement of turbines within an array that makes up a wind farm generating electricity. Can you tell us about the actual process using this evolutionary algorithm methodology and its bio-mimicry characteristics.

Frank: OK, first you start with the area in which you want to place the turbines. You start off with an initial placement. Let’s say that you have 100 turbines that you want to place. In the beginning you might have more than one possible arrangement of the turbines – say there are 20 of them and these 20 potential solutions can be called a population. As in evolution, you would have a process where these ‘parents’ produce ‘children’. The children are another set of possible placements. Depending on whether the children are better than the parents, the children themselves become the new parents for the next iteration. This iteration process continues till, at some point, you come up with a good placement that you are satisfied with.

Matthew: Is a very simple way of describing this process that it’s like a line of best fit?

Frank: What you are trying to do is to improve your solutions over time. The questions of course are how do you construct your next placement and how do you evaluate the quality of your placement of the turbines in the wind farm. For that you need to model the wake effects. Then you optimise according to your model.

Matthew: I’m assuming this is based on your model being constructed around a good real-life scenario. Then you’re running each different test in the lab – you’re not actually placing real turbines in different arrangements in the field.

Frank: What we are currently doing is we take the model, based on the literature, then we optimise it with our algorithms. The next step is to talk to the wind industry, then try to match our algorithms.

Matthew: So what advantages does this particular approach have? What other approaches would people be pursuing? What other theoretical approaches could there be to solve this problem and create, I guess, better bang for your buck, which is what they’re aiming for?

Frank: Because modelling wake effects can become really complex, one of the main advantages is that you have a set or solutions and you can easily parallelize them on different computers, or on different cores of a computer. So you can evaluate each placement in parallel. This is what we are currently doing. So whenever we generate a new set of solutions we parallelize the evaluation, meaning we can look at many solutions in parallel, which reduces the optimisation time, for placements of up to 1000 turbines. If you were to run that on just one computer you would spend something like six months on the whole process. Running the solutions in parallel allows the computing time to reduce from half a year to two weeks.

It’s a really complex optimisation process. Because you have evolutionary processes and can run solutions in parallel, you can use all the super computing power that is currently available.

Matthew: Frank, the German company ENERCON has a 7.5 megawatt turbine now commercially available and SIEMENS has followed with a 6 megawatt turbine. Now you did mention a 1000-turbine wind farm. Is that a likely size for a wind farm once you start talking about 7.5 or 10 megawatt turbines. You’re talking about 7,500 megawatts then. Is that the sort of scale we’re talking about in the North Sea with the shift away from nuclear energy in Germany?

Frank: Well, what we wanted to do in our studies was to try our solutions with 100, 200, 500 and 1000 turbines. We took our setting to the extremes to demonstrate that we can do it for 1000 turbines, in case we do want to have that many. Currently the most likely setting is 200 to 500 turbines. But in principle you could have more and model the wake effects. It’s always a question of design. Also if you want to use the new turbines, that are producing more energy than some years ago, the question is how many turbines do you want to place in a certain area.

Matthew: How long ago have you actually been working on this project? You’re collaborating with MIT. Who actually commenced the project? Did it come out of Max Planck or yourself or the University of Adelaide or the MIT people?

Frank: Well, the MIT people started working on this, I think, a bit earlier. I went over there last year in April but I think they started at the beginning of 2010 or a bit earlier. For the three months I spent there we worked together but now we have extended in different directions. The first initiative came from MIT. Saying that, they want to work on wind energy and to have some impact. Specifically we worked together on the placement of turbines. This is a nice field of research for us from the optimisation point of view.

Matthew: What improvements are you wanting to make to your model? When can we see the work actually play out in the real design and the actual construction and placement of the turbines in real wind farms? When is that likely to occur?

Frank: We are still working on improving our algorithms for minimising the wake effects. The next step is to take into account some other objectives, cable length, for example. What would also be interesting would be to go a bit further and use some of the more complex models of wake effects that are available in the literature, then see how our algorithms go with that.

Talking to the wind industry I think we have some methods that we can plug into simulation software to improve the placement of wind turbines. These methods would be very good to use in the design phase. We have talked to some companies but currently we haven’t found the time to put in a joint project. I think we will see this taking place in the next two or three years.

Matthew: What do you think will be the main outcome? Will you be able to achieve more electricity generated from a constrained site where you don’t have a lot of land, like an island or some highly populated area in Europe? And will you be able to achieve better placement, better value for money in larger field arrays in the ocean? Or are you really just focused on the ocean large array?

Frank: Placing a large number of turbines in a constrained area results in larger wake effects and you lose a lot of energy when there are large wake effects. It’s therefore very valuable to use these algorithms where you are trying to maximise the energy produced from a constrained area.

Matthew: And do you feel that the industry is likely to take on your particular approach? Is that something that is ultimately your goal? Or is this more a theoretical research project?

Frank: I think the wind industry would have their own models of wake effects and some other things. In general our algorithms can also be applied to some other settings. It’s not too complicated to adjust the settings. The design and planning of wind farms is very complex in the beginning so I would think that they would take up these methods.

Matthew: Can you tell us if there any other bio-mimicry projects that you are working on?

Frank: We are using evolutionary algorithms in other areas, not just renewable energy. For example we can apply them to problems in sport. We can apply them to track cycling problems to improve pacing strategy for cyclists. We started on a project on track cycling just three months ago. The other type of research that we are doing is more on the theoretical side, basically to understand how and why they work.

Matthew: Just going back to wind, I was reading an article that suggested that with some of the actual energy that is captured by wind turbines, some effect occurs that people would not intuitively imagine. That is, that higher altitude winds actually come down and power the wind turbines, and therefore that’s a reason for greater spacing. Have you had any exposure to that information?

Frank: No we haven’t.

Matthew: That’s OK. Is there anything else you’d like to cover, that maybe our audience would like to know about wind turbine placement, or even your experience with wind in Germany? For example, how many wind turbines were there where you lived or what’s your personal impression of the view of wind turbines.

Frank: Well the interesting thing about Germany is that it has now decided not to have nuclear power anywhere. So Germany has to increase the capacity for wind energy and other renewable energy sources to a large extent. I think these are interesting times both in Germany and in Australia.

Matthew: Have you heard what the new targets are going to be for solar photovoltaic and wind power? Has the government announced anything as yet?

Frank: Well they have just decided that they will kick out nuclear power a few days ago. So the question is now how to compensate for nuclear power.

Matthew: That’s fantastic. Thank you very much for your time, Frank.

Transcript by Bronwyn L.