Machine Learning Research
NREL's machine learning (ML) research seeks to uncover foundational developments using renewable energy data sets to guide future computational and laboratory experiments.
Detecting Anomalies to Improve Data Center Efficiency
Complex high-performance computing systems are complex and require sophisticated tools for data center efficiency. This work describes a system for analyzing, visualizing, and interpreting large volume streaming data. For more information, see Data Center Facility Monitoring with Physics Aware Approach, High Performance Computing (2023).
Downsampling Large Data Sets
This project proposes an algorithm to select data points from large and high-dimensional data sets to achieve data reduction with minimal information loss. It is achieved by approximating the data probability density function with a specific emphasis on rare data points. It enables efficient data summarization and data-efficient machine learning.
Invertible Neural Network for Airfoils
This work developed a specialized invertible neural network (INN) model that learns a two-way mapping between airfoil shape parameters and aerodynamic/structural performance metrics. The INN framework enables high-fidelity aerodynamic effects to be captured in wind turbine blade design cycles that often require rapid design iterations under changing target specifications. For more information, see Invertible Neural Networks for Airfoil Design, Aeronautics | Astronautics (2022).
Challenges and Opportunities of Machine Learning Control in Building Operations, Building Simulation (2023)
Plug and Play Directed Evolution of Proteins With Gradient-Based Discrete MCMC, Machine Learning: Science and Technology (2023)