European-Brazilian consortium joins forces in advanced HPC wind turbine simulations for use in on- and offshore wind energy farms

12 Oct 2020 Almere - From downtown Almere's virtual studio Primeur Magazinehad the opportunity to interview University of Nottingham coordinator Xuerui Mao and Brazilian partner Bruno Souza Carmo, who works at the University of São Paulo, about the High-Performance Computing for Wind Energy (HPCWE) project. Xuerui Mao explained that HPCWE has 11 partners. Eight of them are from Europe and three are from Brazil. This consortium covers SME industries, HPC centres and universities. The project objectives include promoting the application of HPC in wind energy and promoting the joint collaborations between Europe and Brazil in both academics and industries.

Primeur Magazine: When you talk about HPC in wind energy, what do you mean by that?

Xuerui Mao:In wind energy HPC is a must. In several stages of wind energy, we have to use HPC. For example, when we design the turbine blades, the simulation of flow around the blade is critical. As for the wind farm scale which is larger than the turbine blade scale, we also need HPC to figure out the optimal layout of wind turbines in the farm. In a larger scale such as the 1000-kilometer scale which is also called the meso-scale, similar with the wider scale, we can also run large-scale simulations to predict the overall wind resources so as to pre-construct a wind farm. So, there are several levels in wind energy where we need HPC, from the smallest scale in flow around the blade to the largest scale in the wider scale.

Primeur Magazine: Yes, we saw pictures of it. You basically module everything from the blade and the tip of the blade to large-scale wind energy plants. Is that correct?

Xuerui Mao:Exactly. The scale integration is a critical component of the work. We cover this wide range of scales where HPC resources are required in the development of wind energy.

Primeur Magazine: Then we understand why you need HPC. There are already a lot of simulations, and simulations of parts but you collect everything together, from the small to the large range. This is a European/Brazilian co-working project. Bruno, can you please explain a little bit about this collaboration?

Bruno Souza Carmo:Sure. Brazil is collaborating in this project with the European partners for many reasons. The first of them is the large wind energy potential that we have here in the country. As opposed to what happens in European countries, we still have a lot of onshore potential to be explored. Wind energy exploitation is growing very fast in Brazil. There are many engineering and technology challenges which should be tackled. It would be nice to use high-performance computing to help solve those issues. Besides the onshore capability, we are also starting to develop the offshore capability. We don't have offshore wind farms in Brazil yet but we have many plans to have them constructed very soon.

In terms of the industry, it is very strategic for Brazil to be part of this project. In addition, we also have a growing community in academia, working with wind energy. It would be very strategic for us to also partner with European institutions that have a long-term tradition in this area. We could benefit from this collaboration and also offer some of the work that we are developing here in terms of capacity, people and software. For us it is very interesting to be part of this collaboration. It is a win-win situation for the European and Brazilian partners, both in terms of the academia and industry.

Primeur Magazine: When you look at Europe and Brazil, can we say that in Europe there are more offshore wind parks and in Brazil more onshore farms?

Bruno Souza Carmo:In terms of what is being developed now, yes. A lot of the development in Europe nowadays is offshore because many of the onshore sites are already producing energy which means that there are not so many onshore sites available nowadays. There are some but not a lot. Europe has been developing its offshore potential for quite some years now in the North Sea mainly. There is a huge potential there to be explored. In Brazil we have an offshore potential as well but also some onshore locations that are still waiting to be explored. We have both in Brazil because the wind energy started later to be exploited than in Europe. That is why we are lagging behind in terms of time.

We have already a very clean energy matrix already because we have a lot of hydro power here. So, the pressure for using wind energy was not so strong as in Europe. That is one of the reasons why it wasn't developed so fast but since 10 to 15 years ago we have been investing a lot in wind energy. We see a steady growth of this technology in Brazil.

Primeur Magazine: When you are talking about developing models and using HPC for wind energy, are you then talking mostly about using HPC or are you more focused on developing algorithms?

Xuerui Mao:There are several levels. We use advanced HPC architectures to test our codes. Those facilities are offered by the UK national HPC centre EPCC, hosted by Edinburgh University and also by another partner hosted by the University of Stuttgart. Here, the work is related with hardware architectures. Another level is the development of codes with exascale potentials. For that we identified the codes with such potentials. We identify the Key Performance Indicators (KPIs) to develop those codes for the next generation wind energy prediction tools. At the third level we use some commercial or very mature codes. We modify them and we test them in typical wind farm sites. For example, we have selected one in Europe and another one in Brazil to match the interest of both partners.

Primeur Magazine: The project is now halfway. You recently organized a workshop. Can you tell a little bit about the results that you achieved thus far?

Xuerui Mao:We have developed many advanced algorithms and installed them into several of the codes, the flagship codes, and tested them using the three use cases. For example, for wind resource assessment which is critical for the pre-construction of wind farms, using our advanced technologies we can reduce the prediction error by around fifty percent. Previously, we used the meso-scale simulation to make the prediction. Now, we can have a detailed fluid flow technology which will significantly reduce the computational cost or when we use the same cost, to lift the accuracy significantly. This is one of the examples. Some other examples include the model of the wind turbine and the wind turbine aerodynamics, and structure dynamics which we call aero-elastics. Still other examples are the prediction of uncertainty and qualification of the uncertainty, as well as the optimal layout of wind farms to advance the model of wind turbines and so forth. So, there are several advanced techniques and codes under development.

For the dissemination activities we have presented our work at several conferences and there are several papers accepted and under review. We also held a mid-term project workshop where we have had over 200 attendees. All the presentations, in the form of videos and powerpoints, are available at the HPCWE project website.

Primeur Magazine: The number of participants, 200, sounds like a lot of people working in this area. It was a virtual workshop so everybody could indeed attend. Are there specific topics that were discussed at the workshop which you would like to highlight?

Xuerui Mao:One of the items in our work that we highlighted is the efficient joint simulation of aerodynamics and structural dynamics. We coupled two solvers which was managed by our partner at Imperial College London, joined by the University of São Paulo, in which Dr. Bruno Souza Carmo is leading the work. Another collaboration is between the University of Nottingham and the Technical University of Denmark (DTU). We applied the Weather Research and Forecasting (WRF) data collected from large scale numerical simulations and some limited environments to create data fusion. Those are the techniques we would like to highlight. Of course, there are several more, such as the quantification of uncertainties of wind farm output induced by the oscillation of the free stream wind condition. You know that the wind changes with time strongly. Maybe it can be two meter a second and one hour later it would be 50 meters per second. This would induce the power curve to fluctuate. In this project we developed a framework and applied it to identify such fluctuations and to quantify the uncertainty.

Primeur Magazine: Yes, we are here in Almere and we know a lot about wind changes. Actually, most of the time, we call it a breeze. We have light breezes and strong breezes. We don't call it wind but people who come here, call it wind so we know a little bit about changing winds. Bruno, can you tell a little bit about the work that you do specifically and whether that has specific relations to Brazil or is it just part of the overall work?

Bruno Souza Carmo:In Brazil we have three main partners: the University of São Paulo where I am from; the University of Campinas; and Federal University of Santa Catarina. Each partner has different capabilities. In the case of the University of São Paulo, we are collaborating with Imperial College London, as Xuerui just mentioned, on the modelling and simulation of the turbine at the blade scale, and more specifically on what is called blade-resolved simulations, and also using actuator line modelling to perform simulations at the wind farm scale. All of this is called micro-scale by the wind energy community.

The people from Unicamp, the University of Campinas, work more with optimization and sensitivity calculation of the wind farm layout. The capabilities are more focused on the optimization part of the project. The Federal University of Santa Catarina team has been working on the integration between meso- and micro-scale and also providing some field data to do some studies in terms of a possible offshore location for a wind farm in Brazil. They have a team there with some lighters, those sensors that measure the speed of the wind on different heights. We are providing this data and we can test the models that have been developed in the project. We have different institutions with different capabilities and we are trying to make the best matches between the European and Brazilian partners to get the best value out of the project.

Primeur Magazine: It looks like there is already a lot of progress. What are the things that you will do in the coming year of the project?

Xuerui Mao:In the next year our top priority is to focus on the codes with exascale potential and try to develop them. Those codes are firstly developed in Europe. Apart from the exacale potential, they also have good capacities and good applications in wind energy for the three use cases we have defined in this project. The development of exascale codes is our first priority in the next one year while some other parts include further development of Big Data analytical scales. For example, we want to reduce the data storage. In wind energy, for example when we want to construct a farm, the wind may come from various directions in each season. We want to test the wind in 10 years. This means you have a huge amount of data to store. We want to develop a technology to reduce this data. We want to extract only the key dynamics, the key models of this complex flow. If we have a mountainous region, the flow is really complicated. It is three-dimensional. We want to use the Big Data analytical skills we are developing to reduce the size of the data so as to facilitate our project and to reduce the data to save during the simulations.

Bruno Souza Carmo:At the Brazilian side, many of the developments that were carried out in the first year were like concept tests. We had to put the team working together and do some benchmark tests. Now, the second year will be more about production in terms of taking it to the exascale, as Xuerui just mentioned. We will try to solve a very complicated case. I see the development of the project as a whole: the first year is about preparation of things and then the second year is about production and use of the tools that have been developed for use cases that are at industry layer level.

Primeur Magazine: We are really looking forward to see the results of the project and we hope to talk to you both again. Thank you very much.

The HPCWE participants include the following partners:

1. European Universities: University of Nottingham (coordinator), Imperial College London, University of Twente, Technical University of Denmark

2. European HPC centres: HLRS at the University of Stuttgart, EPCC at the University of Edinburgh

3. European wind industries: EDF, Vortex (SME)

4. Brazilian partners: University of São Paulo, University of Campinas, Federal University of Santa Catarina

The scientific advisory board includes Vestas (Europe wind industry), Case dos Ventos (Brazil wind industry), and ITA (Brazil institute).

More information is available at the HPCWE project website. The presentations of the workshop are available online .

Ad Emmen