David A. Romero | Searcharxiv

arXiv2024

Deep encoder-decoder hierarchical convolutional neural networks for conjugate heat transfer surrogate modeling

Conjugate heat transfer (CHT) analyses are vital for the design of many energy systems. However, high-fidelity CHT numerical simulations are computationally intensive, which limits their applications such as design optimization, where hundreds to thousands of evaluations are required. In this work, we develop a modular deep encoder-decoder hierarchical (DeepEDH) convolutional neural network, a novel deep-learning-based surrogate modeling methodology for computationally intensive CHT analyses. Leveraging convective temperature dependencies, we propose a two-stage temperature prediction architecture that couples velocity and temperature fields. The proposed DeepEDH methodology is demonstrated by modeling the pressure, velocity, and temperature fields for a liquid-cooled cold-plate-based battery thermal management system with variable channel geometry. A computational mesh and CHT formulation of the cold plate is created and solved using the finite element method (FEM), generating a dataset of 1,500 simulations. Our performance analysis covers the impact of the novel architecture, separate DeepEDH models for each field, output geometry masks, multi-stage temperature field predictions, and optimizations of the hyperparameters and architecture. Furthermore, we quantify the influence of the CHT analysis' thermal boundary conditions on surrogate model performance, highlighting improved temperature model performance with higher heat fluxes. Compared to other deep learning neural network surrogate models, such as U-Net and DenseED, the proposed DeepEDH architecture for CHT analyses exhibits up to a 65% enhancement in the coefficient of determination $R^{2}$. (*Due to the notification of arXiv "The Abstract field cannot be longer than 1,920 characters", the appeared Abstract is shortened. For the full Abstract, please download the Article.)

cs.CE↗

arXiv2021

Optimizing wind farms layouts for maximum energy production using probabilistic inference: Benchmarking reveals superior computational efficiency and scalability

Successful development of wind farms relies on the optimal siting of wind turbines to maximize the power capacity under stochastic wind conditions and wake losses caused by neighboring turbines. This paper presents a novel method to quickly generate approximate optimal layouts to support infrastructure design decisions. We model the quadratic integer formulation of the discretized layout design problem with an undirected graph that succinctly captures the spatial dependencies of the design parameters caused by wake interactions. On the undirected graph, we apply probabilistic inference using sequential tree-reweighted message passing to approximate turbine siting. We assess the effectiveness of our method by benchmarking against a state-of-the-art branch and cut algorithm under varying wind regime complexities and wind farm discretization resolutions. For low resolutions, probabilistic inference can produce optimal or nearly optimal turbine layouts that are within 3% of the power capacity of the optimal layouts achieved by state-of-the-art formulations, at a fraction of the computational cost. As the discretization resolution (and thus the problem size) increases, probabilistic inference produces optimal layouts with up to 9% more power capacity than the best state-of-the-art solutions at a much lower computational cost.

math.OC↗