Long Short Term Memory Network for Predicting Refrigerant Dynamic Behaviors in Air Conditioning System

DOI Web Site
  • MIYAWAKI Kosuke
    Advanced Technology Center, Mitsubishi Electric Co.,
  • YAMAGISHI Rena
    Living Environment Systems Laboratory, Mitsubishi Electric Co., Institute of Industrial Science, The University of Tokyo
  • SCIAZKO Anna
    Institute of Industrial Science, The University of Tokyo
  • SHIKAZONO Naoki
    Institute of Industrial Science, The University of Tokyo

Bibliographic Information

Other Title
  • 空調機の冷媒分布予測のための長・短期記憶ネットワークの実用性評価
  • クウチョウキ ノ レイバイブンプ ヨソク ノ タメ ノ ナガ ・ タンキ キオク ネットワーク ノ ジツヨウセイ ヒョウカ

Search this article

Description

<p>Long short term memory network was applied to predict the dynamic refrigerant behaviors in an air conditioning system. The transitions of refrigerant distributions in four elements were predicted under different compressor starting speeds and charging amounts of the refrigerant. The network can predict the behavior of refrigerant distribution precisely within 4% error when a sufficient number of training data is used. Although the lack in training data leads to prediction error of 8.5%, introduction of conventional design knowledge of the refrigerant system to the network enables to improve the prediction accuracy up to 5.4% error.</p>

Journal

Keywords

Details 詳細情報について

Report a problem

Back to top