• Search Research Projects
  • Search Researchers
  • How to Use
  1. Back to previous page

モデル予測制御と強化学習を応用したビルエネルギーシステムの運用最適化手法の開発

Research Project

Project/Area Number 23KJ0513
Research Category

Grant-in-Aid for JSPS Fellows

Allocation TypeMulti-year Fund
Section国内
Review Section Basic Section 23020:Architectural environment and building equipment-related
Research InstitutionKyushu University

Principal Investigator

GAO YUAN  九州大学, カーボンニュートラル・エネルギー国際研究所, 助教

Project Period (FY) 2023-04-25 – 2025-03-31
Project Status Discontinued (Fiscal Year 2023)
Budget Amount *help
¥1,800,000 (Direct Cost: ¥1,800,000)
Fiscal Year 2024: ¥900,000 (Direct Cost: ¥900,000)
Fiscal Year 2023: ¥900,000 (Direct Cost: ¥900,000)
KeywordsReinforcement learning / Deep learning / Net zero energy / Renewable energy
Outline of Research at the Start

This study uses the deep learning model to improve the control accuracy of the MPC algorithm and achieves the best control effect through the combination of the two prediction models. Through our research, it is believed that the automatic control of the energy system can be realized.

Outline of Annual Research Achievements

The research content mentioned in the application topic has been essentially completed to a high standard. Based on the renewable building energy system that actually exists in Tsukuba, Ibaraki Prefecture, Japan, we explored various possibilities of employing Model Predictive Control and Reinforcement Learning algorithms for its operational control. Deep learning models play a significant role in the prediction of renewable energy systems. The initial concern involves the interpretability of deep learning applications. When employing deep learning predictive models in renewable energy systems, the interpretability of these models becomes crucial. Consequently, system owners need to evaluate the computational process of the model and make informed final decisions.

Current Status of Research Progress
Current Status of Research Progress

1: Research has progressed more than it was originally planned.

Reason

The applicant has basically completed research on actual renewable energy systems, and has published a total of 12 high-level papers in related fields in Applied Energy, Energy and other journals, all of which the applicant is the first author.

The cooperative company has begun implementation testing of the algorithm.

Strategy for Future Research Activity

The next aspect concerns the practical deployment and application of fault diagnosis models within the system. Our study, using data from air conditioning systems as an example, aims to reduce the parameter count of the proposed model without compromising the accuracy of fault diagnosis. This approach is intended to enhance computational speed in engineering applications.

In summary, the applicant has successfully achieved the established objectives set forth in the application topic and has also produced numerous scholarly papers.

In the future, we will further improve the efficiency of the algorithm based on previous research and conduct more actual case analysis and implementation.

Report

(1 results)
  • 2023 Research-status Report
  • Research Products

    (6 results)

All 2024 2023

All Journal Article (6 results) (of which Int'l Joint Research: 6 results,  Peer Reviewed: 6 results)

  • [Journal Article] Successful application of predictive information in deep reinforcement learning control: A case study based on an office building HVAC system2024

    • Author(s)
      Gao Yuan、Shi Shanrui、Miyata Shohei、Akashi Yasunori
    • Journal Title

      Energy

      Volume: 291 Pages: 130344-130344

    • DOI

      10.1016/j.energy.2024.130344

    • Related Report
      2023 Research-status Report
    • Peer Reviewed / Int'l Joint Research
  • [Journal Article] Adversarial discriminative domain adaptation for solar radiation prediction: A cross-regional study for zero-label transfer learning in Japan2024

    • Author(s)
      Gao Yuan、Hu Zehuan、Shi Shanrui、Chen Wei-An、Liu Mingzhe
    • Journal Title

      Applied Energy

      Volume: 359 Pages: 122685-122685

    • DOI

      10.1016/j.apenergy.2024.122685

    • Related Report
      2023 Research-status Report
    • Peer Reviewed / Int'l Joint Research
  • [Journal Article] Interpretable deep learning for hourly solar radiation prediction: A real measured data case study in Tokyo2023

    • Author(s)
      Gao Yuan、Miyata Shohei、Akashi Yasunori
    • Journal Title

      Journal of Building Engineering

      Volume: 79 Pages: 107814-107814

    • DOI

      10.1016/j.jobe.2023.107814

    • Related Report
      2023 Research-status Report
    • Peer Reviewed / Int'l Joint Research
  • [Journal Article] Automated fault detection and diagnosis of chiller water plants based on convolutional neural network and knowledge distillation2023

    • Author(s)
      Gao Yuan、Miyata Shohei、Akashi Yasunori
    • Journal Title

      Building and Environment

      Volume: 245 Pages: 110885-110885

    • DOI

      10.1016/j.buildenv.2023.110885

    • Related Report
      2023 Research-status Report
    • Peer Reviewed / Int'l Joint Research
  • [Journal Article] Spatio-temporal interpretable neural network for solar irradiation prediction using transformer2023

    • Author(s)
      Gao Yuan、Miyata Shohei、Matsunami Yuki、Akashi Yasunori
    • Journal Title

      Energy and Buildings

      Volume: 297 Pages: 113461-113461

    • DOI

      10.1016/j.enbuild.2023.113461

    • Related Report
      2023 Research-status Report
    • Peer Reviewed / Int'l Joint Research
  • [Journal Article] How to improve the application potential of deep learning model in HVAC fault diagnosis: Based on pruning and interpretable deep learning method2023

    • Author(s)
      Gao Yuan、Miyata Shohei、Akashi Yasunori
    • Journal Title

      Applied Energy

      Volume: 348 Pages: 121591-121591

    • DOI

      10.1016/j.apenergy.2023.121591

    • Related Report
      2023 Research-status Report
    • Peer Reviewed / Int'l Joint Research

URL: 

Published: 2023-04-26   Modified: 2024-12-25  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi