Engineering speed modelling of realistic fatigue for all the individual turbines in wind parks by representative pre-calculations

Wind farms are very complicate systems. A farm consists of many turbines which individual behavior is dependent on the weather and the other turbines in the farm. The wear of tear on different turbines in a farm can vary allot. Some turbines can for example be worn out after 20 years, while others have many years of remaining useful lifetime. Simulating wind farms is difficult for practical industrial use. You could use a very advanced wind farm simulator that does a good job at predicting the behavior of each turbine. This can however require so much computational resources and time that it is unfeasible for practical use. Another way is to use much simpler simulation tools that are fast enough for practical use, but they are however less realistic. We have a novel approach in the our NEXTFARM project that we believe can solve the dilemma. We attack the problem from the perspective of the individual turbine. It only cares about the wind that hit it, and not how it came about. If we can describe all the wind conditions any turbines in a farm can experience, then we can feed this to wind turbine simulator and realistically predict its behavior for any condition. Many thousand wind turbine simulations are needed to cover all eventualities, but its only done once. We can then use a fast wind farm model and replace the highly simplified wind turbine behavior such simulators use with our pre-calculated simulation results. Such a wind farm simulator could simulate the whole lifetime of a wind farm in minutes. It will enable engineers to design cheaper new wind farm. It will enable wind farm operators with an existing farm to run it longer and more efficiently. Both will contribute to reducing the cost of renewable and environmentally friendly electricity from wind farms on land or offshore.

The project conducted perhaps the most thorough benchmarking of leading Dynamic Wake Meandering (DWM) models and high-fidelity Computational Fluid Dynamics (CFD) wind farm models ever undertaken. IFE and NTNU utilized the developed DWM wind farm model, WIFET Farm. DTU deployed their DWM model, while NREL used their FAST.Farm model. Uppsala University conducted the supercomputer CFD simulations. The models were compared with respect to power, loads, and fatigue. The DWM models were also compared with detailed full-scale measurements from the Lillgrund offshore wind farm, provided by Vattenfall and Siemens Gamesa. This work made it possible to identify and address the strengths and weaknesses of the various choices of sub-models that all DWM models are built from. This, in turn, has advanced the DWM wind farm wake modeling research field to a new level of realism. The project developed a new DWM model, WIFET Farm, which is easy to use, fast, and well-suited for industry use. It performed well compared to the other DWM models. The project has significantly advanced the international DWM wind farm modeling research field. International leaders in the field, DTU and NREL, have planned or started work based on the project's findings. In Norway, IFE has built competence that places it at the forefront of international research in DWM wind farm modeling. DWM models are arguably the most realistic yet practical tools for industry purposes. They can improve the economy of wind farms in the planning and engineering phase, during operation, and for lifetime planning and extension. The project has produced improved competence and tools that are now available to the industry, enhancing the industry's understanding of DWM wind farm modeling. Improved DWM knowledge and tools will reduce the cost of wind energy, which holds significant value and importance for society.

The project will be divided in three phases, In the first part a simplified and parameter-based version of a wake model will be developed, in order to achieve engineering friendly computational speeds for our turbine inflow model, a reduced version of the Dynamic Meandering Wake model (DWM) will be developed. In particular, the problem will be treated from the point of view of the single turbine, rather than of the whole wind park. The main assumption is that, deep within the interacting wind park, a turbine will experience only the wakes of the closest upstream turbines as concentrated, meandering wakes. This means concentrated wakes will be shed by turbines from maximum one or two rows upstream, positioned in a certain angular sector, aligned with the wind direction. The wakes of the turbines further upstream will be experienced only as a diffused wakes, generating a deep wind park velocity deficit. In the second part of the project, direct correlations between fatigue and local inflow for each turbine will be evaluated, studied, developed. This will be done either by simulating a large number of load cases and evaluating the resultant matrix. Advanced algorhitms for complex systems and matrix reduction will be studied in order to reduce the resultant matrix and generate a look up table that can be used to directly estimate fatigue from local inflow condition. The resultant model will then be successively verified and validated against other similar models and against full scale available data. In the third phase of the project, the developed tools and models will be applied to cost models and their impact on the wind park LCOE of a farm will be studied