In recent years, there has been a growing interest on sustainable energy resources, such as solar, geothermal, wave, and wind energy. The renewable energy sector’s share of the energy supply is expected to grow to 18.6% by 2030. Moreover, the International Energy Agency (IEA) predicts that wind energy will have a 12% share of the global energy supply by 2050. To reach this target, the wind energy production capacity will have to increase at an average rate of 47 GW/year, resulting in massive investments and a dynamic, growing market for wind-related technology and services.

In this context, the overall vision for this aspect of our research program is to leverage modeling, simulation and optimization methods to optimize energy systems by maximizing their energy efficiency, minimizing their cost, and minimizing their environmental impact. The following are two application-specific projects that we are currently pursuing:

Wind Farms

Accelerating wind farm power assessment with neural networks

We are developing a methodology for fast, accurate prediction of the flow field and power generation in wind farms, using Deep Convolutional Hierarchical Encoder-Decoder Neural Networks with an architecture as depicted in the figure below. With this approach, we leverage computationally efficient software platforms that have been developed for image-based neural networks to realize near-real-time predictions of the power output of wind farms, thus enabling interactive design of wind turbine layouts.

Encoder-decoder architecture for image-based prediction of the flow field within wind farms.

We have achieved fast prediction of power generation in wind farms with less than 1% error with respect to state-of-the-art CFD simulations, as shown in the figures, even when the neural networks are trained with relatively small datasets.

Prediction errors for wind speed and power output as a function of the size of the dataset used to train the networks. Left: Baseline architecture, Right: Optimized architecture (depicted in the previous figure).

Simulation and design optimization of wind turbine layouts

Our research also focuses on modeling wind turbine wakes using CFD methods. The turbulent nature of turbine wakes and the complex geometries of turbine blades increase modeling challenges. In the absence of high-quality experimental data of multiple wake interactions in complex terrains, CFD simulations can provide valuable insight in these complex flow phenomena. One of the main goals in this project is to develop a relatively low cost CFD model without loss of physical representation. Such models enable simulation-based optimization of the turbine layout with adjoint methods, leading to the design of wind farms that take advantage of both the stochastic wind resource profile and local terrain effects to maximize their energy generation.