Role of Artificial Intelligence in Large Wind Turbine Designs

The utilization of wind energy across various sectors has increased significantly in recent decades. It is considered one of the most promising clean energy sources and has matured in terms of technological readiness and commercial penetration. In fact, wind energy has experienced the fastest growth compared to other renewable energy sources. Wind energy has clearly become an essential part of human life worldwide and has undergone massive development in various areas [1]. This growth is driven by the ongoing decline of fossil fuel reserves, concerns about a potential energy crisis, energy price volatility, and the growing public awareness of environmentally friendly energy alternatives. There is concern that the ongoing effects of climate change will have significant consequences on ecosystems, healthcare systems, and the general economy in the coming decades [2]. One of the key success factors in wind energy adoption is its maturity in all aspects: research, design, manufacturing, logistics, operation, maintenance, market openness, and regulation. Wind turbine technology is multidisciplinary and has been continuously tested over decades of operation.

As the use of wind energy continues to grow rapidly, it presents new challenges that must be addressed. One major challenge in wind energy development is the availability of space, especially as renewable energy demand continues to rise. In response to this, wind turbine sizes have been increasing considerably in the last decades. Wind turbines have now scaled up to the 10–15 MW range, with rotor diameters exceeding 200 m, and are expected to reach 20+ MW with rotor diameters of over 300 m in the near future. This evolution introduces a whole new set of challenges compared to earlier turbine technologies. For instance, ultra-large wind turbines feature much longer, more slender, and more flexible blades. As the wind turbine size increases, various issues begin to emerge, with one of the most critical being the increased susceptibility to blade instability. Wind turbine technological advancements must keep pace to ensure the safety and structural integrity of these large machines. This directly implies that improvements are needed across multiple areas of wind turbine technology.

Artificial intelligence (AI) has experienced significant growth in its application across various areas in recent years. Its adoption has been remarkably rapid, and the wind energy community is increasingly recognizing the potential of AI. This can be applied to wind turbine designs [3,4,5], enhancing controller robustness [6,7], improving prediction tools and site assessments [8,9,10,11], optimizing/scheduling maintenance [12,13], and managing energy storage systems [14,15,16].

There is a clear trend toward the adoption of AI for advancing wind turbine technology, suggesting that AI could play an important role in the future success story of wind energy. From the brief analysis above, it becomes evident that while large wind turbines certainly pose significant challenges, they also offer substantial opportunities. A key question for the future is: Can AI be effectively leveraged in the development of large wind turbines? This editorial specifically aims to provide a concise overview of AI’s role in large wind turbine designs and to highlight its potential future directions.

In the wind energy sector, AI has naturally found its place—primarily in optimization. This is largely because optimization techniques have been employed in the wind energy field for decades [17,18,19,20]. Therefore, as AI becomes a norm, its integration into the development workflow is a logical progression. Wind turbine and controller designs are typically developed using highly specialized software, often carried out by different companies. For example, while the wind turbine blades and tower might be designed by the original equipment manufacturer (OEM), substructures such as monopiles, jackets, or floating foundations are usually designed by separate companies. One potential role of AI is in improving the protection of intellectual property (IP) across the various stages of the design workflow. Safely sharing these models while protecting the IP remains one of the most challenging bureaucratic tasks in the workflow. One idea is to use AI to replace the turbine structures during the substructure design process. Another potential application is by employing AI in encryption technologies to securely share these models.

AI may also be incorporated to enhance the prediction accuracy for large wind turbines. There is a particular concern regarding the instability of long and flexible blades. Accurate prediction of aerodynamic damping has proven to be vital for reliably estimating the wind turbine loads progression over time [21,22,23]. Progress has been made in predicting instability in wind turbine blades through the development of advanced dynamic stall models, such as the IAG model [24]. This model works well in the common post stall regime at angles of attack ($\alpha$) around ±30° (denoted as stall-induced vibration zone 1, SIV 1), as well as in the reversed flow regime at $\alpha$ around ±160° (referred to stall-induced vibration zone 2, SIV 2). However, there remains gap in engineering modeling when the blade is perpendicular with the inflow at $\alpha$ about ±90° (known as vortex-induced vibration zone, VIV). Performing computational fluid dynamics (CFD) for VIV is extremely expensive and not straightforward, which makes it difficult to recreate instability cases. AI could potentially fill this gap by bridging the prediction capabilities between SIV and VIV in engineering models for wind turbines. One promising method is to train AI models using field data to represent various VIV scenarios. AI may also be employed to uncover relationships between spanwise vortex movement during VIV. Further applications of AI include improving engineering sub-correction models, such as three-dimensional blade root polar corrections or modeling approaches for highly loaded rotor cases.

Site-specific wind turbine design workflows may also benefit from AI utilization. There are already documented activities concerning this matter, for instance, using AI to reconstruct wind turbine wakes from sparse sensor data [9,11,25] or wind farm layout analysis [26,27]. Future directions of projects could be aimed at bridging the gap between wind farm level analysis and individual wind turbine loads. Design loads cases (DLCs) governed by standards [28,29] mainly regulate loads assessments for individual wind turbines. However, the impacts of a wind farm on turbine loads are not yet evaluated in great detail [30]. In addition, it is known that generating turbulent wind files is not a trivial task, and matching them with measured timeseries data from the field is even more challenging, especially when accounting for upstream wind turbine wake effects. Based on limited timeseries data, AI could provide the necessary ingredients to reconstruct turbulent wind fields. This can be further adopted in DLC analyses that consider the influence of upstream turbine wakes in a wind farm.