Researchers at the U.S. Department of Energy's Pacific Northwest National Laboratory (PNNL) have developed a method for designing automated control systems that incorporate advances in deep learning and control theory.
Said PNNL's Aaron Tuor, "What we are trying to do is bring this deep-learning–based modeling into a more data-efficient regime enabling its use in real-world applications, which may need interpretability and guarantees of operation that black-box deep-learning modeling can't offer."
The researchers' hybrid approach aims to embed the known and learn the unknown physics of the system to be controlled.
They applied this approach to ordinary differential equations and were able to model a differential equation as a deep neural network.
They used the technique to model and control a building's thermal system, and found that solutions with domain knowledge embedded in the structure of the neural network performed the best.
From Pacific Northwest National Laboratory
View Full Article
Abstracts Copyright © 2020 SmithBucklin, Washington, DC, USA