Language selection

Government of Canada / Gouvernement du Canada

Search

Optimizing thermodynamic trajectories using evolutionary and gradient-based reinforcement learning

From National Research Council Canada

DOIResolve DOI: https://doi.org/10.1103/PhysRevE.104.064128
AuthorSearch for: 1ORCID identifier: https://orcid.org/0000-0002-8603-3027; Search for: ORCID identifier: https://orcid.org/0000-0002-0522-4148; Search for: ; Search for: ORCID identifier: https://orcid.org/0000-0003-1768-6873; Search for: ; Search for: ORCID identifier: https://orcid.org/0000-0002-8146-6667
Affiliation
  1. National Research Council of Canada. Security and Disruptive Technologies
FunderSearch for: Natural Sciences and Engineering Research Council of Canada; Search for: U.S. Department of Energy; Search for: Office of Science; Search for: Basic Energy Sciences
FormatText, Article
Abstract
Using a model heat engine, we show that neural-network-based reinforcement learning can identify thermodynamic trajectories of maximal efficiency. We consider both gradient and gradient-free reinforcement learning. We use an evolutionary learning algorithm to evolve a population of neural networks, subject to a directive to maximize the efficiency of a trajectory composed of a set of elementary thermodynamic processes; the resulting networks learn to carry out the maximally efficient Carnot, Stirling, or Otto cycles. When given an additional irreversible process, this evolutionary scheme learns a previously unknown thermodynamic cycle. Gradient-based reinforcement learning is able to learn the Stirling cycle, whereas an evolutionary approach achieves the optimal Carnot cycle. Our results show how the reinforcement learning strategies developed for game playing can be applied to solve physical problems conditioned upon path-extensive order parameters.
Publication date
PublisherAmerican Physical Society
In
  • Physical Review E 104, no. 6, 064128 (December 2021).
LanguageEnglish
Peer reviewedYes
Export citationExport as RIS
Report a correctionReport a correction (opens in a new tab)
Record identifierfe303482-b589-4345-b834-6728915621a1
Record created2023-03-27
Record modified2023-03-27
Date modified: