Analyzing real-world engineering problems such as wake modeling of wind/ocean current turbines are known to be complex and challenging. The multivariable nature of these problems requires either the implementation of computational analyses under certain simplifying assumptions or conducting experiments for a limited number of scenarios. Hence, there is always a number of key features missed in understanding the key players in determining the energy density, cost of energy, etc.

Machine learning (ML) algorithms have been shown to be a promising solution to complement the existing physics-based wake modeling techniques. Conventional ML algorithms such as statistical models require long-term historical data; this is either not feasible in many real-case applications or it becomes very expensive and time-consuming to collect the required data. Moreover, there are often infinite features in datasets with complex relationships between them. It makes the tasks of feature selection and model tuning become more challenging. There is a need to migrate from such discriminative ML models to generative models to take over the incapability of former models in representing relations between features and classes in the dataset. The generative models offer a rich representation of the independence relation in the dataset.

In this work, a cross-domain study of physics and generative models (e.g. Generative Adversarial Networks) will be performed to show the relationship between the underlying structure of the physical features and ML model. The key achievement of this work is two-fold: first, suggesting a group of emerging generative models in the wake modeling domain; second, reducing the computational cost by demanding smaller dataset. Finally, sensitivity analyses on the impact of this multidisciplinary study on the wind farm capacity factor and cost of energy will further highlight the advantages of this methodology.


A Take on Wake Modeling of Turbines Based on Deep Learning

Division / Sponsors: 
Power Division & Nuclear Engineering Division (NED)

Power Conference
Paper Proceedings: https://asmedigitalcollection.asme.org/POWER/POWER2020/volume/83747

International Conference on Nuclear Engineering
Paper Proceedings: https://asmedigitalcollection.asme.org/ICONE/ICONE2020/volume/83761

POWER-ICONE - Virtual Conference

Supporting the need among researchers worldwide for additional online and multimedia content, ASME now offers the ASME Conference Video Collection. This platform will collect and provide long-term access to conference presentation videos, encompassing all technology areas within ASME's technical conference and events program. About 30,000-40,000 engineers across all mechanical engineering and related disciplines participate in ASME conferences each year. This includes authors of approximately 6,000 to 8,000 technical papers published by ASME, the majority of whom also submit video presentations for our virtual conferences. The video collection will be available as an add-on to existing product packages or as a separate subscription.