Near-optimal Deep Reinforcement Learning Policies from Data for Zone Temperature Control

被引:1
|
作者
Di Natale, Loris [1 ,2 ]
Svetozarevic, Bratislav [1 ]
Heer, Philipp [1 ]
Jones, Colin N. [2 ]
机构
[1] Urban Energy Syst Lab, Swiss Fed Labs Mat Sci & Technol Empa, CH-8600 Dubendorf, Switzerland
[2] Swiss Fed Inst Technol Lausanne EPFL, Lab Automat, CH-1015 Lausanne, Switzerland
基金
瑞士国家科学基金会;
关键词
D O I
10.1109/ICCA54724.2022.9831914
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
Replacing poorly performing existing controllers with smarter solutions will decrease the energy intensity of the building sector. Recently, controllers based on Deep Reinforcement Learning (DRL) have been shown to be more effective than conventional baselines. However, since the optimal solution is usually unknown, it is still unclear if DRL agents are attaining near-optimal performance in general or if there is still a large gap to bridge. In this paper, we investigate the performance of DRL agents compared to the theoretically optimal solution. To that end, we leverage Physically Consistent Neural Networks (PCNNs) as simulation environments, for which optimal control inputs are easy to compute. Furthermore, PCNNs solely rely on data to be trained, avoiding the difficult physics-based modeling phase, while retaining physical consistency. Our results hint that DRL agents not only clearly outperform conventional rule-based controllers, they furthermore attain near-optimal performance.
引用
收藏
页码:698 / 703
页数:6
相关论文
共 50 条
  • [31] Near-Optimal Offline Reinforcement Learning via Double Variance Reduction
    Yin, Ming
    Bai, Yu
    Wang, Yu-Xiang
    ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 34 (NEURIPS 2021), 2021, 34
  • [32] The benefit of receding horizon control: Near-optimal policies for stochastic inventory control
    Dural-Selcuk, Gozdem
    Rossi, Roberto
    Kilic, Onur A.
    Tarim, S. Armagan
    OMEGA-INTERNATIONAL JOURNAL OF MANAGEMENT SCIENCE, 2020, 97
  • [33] Near-optimal reinforcement learning framework for energy-aware sensor communications
    Pandana, C
    Liu, KJR
    IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, 2005, 23 (04) : 788 - 797
  • [34] Near-Optimal Provable Uniform Convergence in Offine Policy Evaluation for Reinforcement Learning
    Yin, Ming
    Bai, Yu
    Wang, Yu-Xiang
    24TH INTERNATIONAL CONFERENCE ON ARTIFICIAL INTELLIGENCE AND STATISTICS (AISTATS), 2021, 130
  • [35] Autoregressive Policies for Continuous Control Deep Reinforcement Learning
    Korenkevych, Dmytro
    Mahmood, A. Rupam
    Vasan, Gautham
    Bergstra, James
    PROCEEDINGS OF THE TWENTY-EIGHTH INTERNATIONAL JOINT CONFERENCE ON ARTIFICIAL INTELLIGENCE, 2019, : 2754 - 2762
  • [36] AutoDDL: Automatic Distributed Deep Learning With Near-Optimal Bandwidth Cost
    Chen, Jinfan
    Li, Shigang
    Guo, Ran
    Yuan, Jinhui
    Hoefler, Torsten
    IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS, 2024, 35 (08) : 1331 - 1344
  • [37] COMPUTATION OF NEAR-OPTIMAL CONTROL POLICIES FOR FREE-RADICAL POLYMERIZATION REACTORS
    OSAKADA, K
    FAN, LT
    JOURNAL OF APPLIED POLYMER SCIENCE, 1970, 14 (12) : 3065 - &
  • [38] Dynamic compensator-based near-optimal control for unknown nonaffine systems via integral reinforcement learning
    Lin, Jinquan
    Zhao, Bo
    Liu, Derong
    Wang, Yonghua
    NEUROCOMPUTING, 2024, 564
  • [39] Optimal, near-optimal, and robust epidemic control
    Dylan H. Morris
    Fernando W. Rossine
    Joshua B. Plotkin
    Simon A. Levin
    Communications Physics, 4
  • [40] Near-optimal energy management for plug-in hybrid fuel cell and battery propulsion using deep reinforcement learning
    Wu, Peng
    Partridge, Julius
    Anderlini, Enrico
    Liu, Yuanchang
    Bucknall, Richard
    INTERNATIONAL JOURNAL OF HYDROGEN ENERGY, 2021, 46 (80) : 40022 - 40040