Researchers Use Convolutional Neural Networks to Predict Airflow Characteristics in the Aerodynamic Profiles of High-Power Wind Turbines and Reduce Computational Time in Four Orders of Magnitude

Wind energy has become a notable source of electricity generation to create a cleaner and more sustainable energy paradigm. However, the performance of wind turbines today needs to be improved if they compete with traditional energy sources. Installing flow control devices on airfoils is one method for achieving this. These flow control mechanisms are installed on the airfoils to increase the wind turbine rotors’ aerodynamic efficiency. The most widely used technique for examining these devices is computational fluid dynamics (CFD) simulations. With the development of artificial intelligence (AI) in recent years, forecasting flow characteristics using neural networks have gained popularity. A team of researchers at the Vitoria-Gasteiz Faculty of Engineering of the UPV/EHU-University of the Basque Country has employed convolutional neural networks to forecast airflow characteristics in the aerodynamic profiles of high-power wind turbines. They have demonstrated that it is possible to study flow control systems using these neural networks, with tolerable errors and a four-order-of-magnitude decrease in calculation time. The study has also been published in Scientific Reports.

The cost per megawatt hour is lower because the same wind turbine can generate more megawatts, which reduces the implementation cost to nearly nothing. The CFD software can imitate fluid movement, which demands a lot of processing power, such as high-speed computers and computation time. However, as previously mentioned, neural networks can forecast flow characteristics. The team also worked on implementing a convolutional neural network (CNN), which chooses several variables utilized for wind turbine flow control. The findings demonstrate that, with minimal errors, the CNN proposed for field prediction can reliably predict the primary flow features around the flow control device.

Regarding the aerodynamic coefficients, the suggested CNN can accurately predict both the trend and the values. Additionally, compared to CFD simulations, using CNNs lowers computational time by a factor of four. The model generates results almost instantly, with some scenarios having an acceptable error range of 5–6%. Two distinct flow control devices (rotating microtabs and Gurney flaps) have been used to initiate the CFD simulations. This served as the output information that the CNN was trained on. The geometry is used as the input in the process, and the CFD results are used as the output. By doing this, one can be sure that the network has been adequately trained and can anticipate the new velocity and pressure fields even when the geometry is different. 

The researchers believe they have created a quick, adaptable, and affordable tool that the industry needs. The current situation calls for a versatile tool that can be used with any aerodynamic airfoil, device system, and even different geometries. The development of the model must take into account the advancement in AI if it is to thrive in the current, cutthroat industrial climate, particularly in the global marketplaces.

This Article is written as a research summary article by Marktechpost Staff based on the research paper 'CNN‑based fow control device
modelling on aerodynamic airfoils'. All Credit For This Research Goes To Researchers on This Project. Check out the paper and reference article.

Please Don't Forget To Join Our ML Subreddit

Khushboo Gupta is a consulting intern at MarktechPost. She is currently pursuing her B.Tech from the Indian Institute of Technology(IIT), Goa. She is passionate about the fields of Machine Learning, Natural Language Processing and Web Development. She enjoys learning more about the technical field by participating in several challenges.