化工进展 ›› 2021, Vol. 40 ›› Issue (2): 664-670.DOI: 10.16085/j.issn.1000-6613.2020-0600

• 化工过程与装备 • 上一篇    下一篇

基于灰色关联BP神经网络的压缩式蓄冷系统中的水合物生成量预测

杨文宇(), 谢应明(), 闫坤, 邹俊华, 舒胜   

  1. 上海理工大学能源与动力工程学院,上海 200093
  • 收稿日期:2020-04-17 出版日期:2021-02-05 发布日期:2021-02-09
  • 通讯作者: 谢应明
  • 作者简介:杨文宇(1994—),男,硕士研究生,研究方向为气体水合物。E-mail:1348485399@qq.com
  • 基金资助:
    国家自然科学基金(50806050)

Prediction of hydrate production in compressive cold storage system based on grey relational BP neural network

Wenyu YANG(), Yingming XIE(), Kun YAN, Junhua ZOU, Sheng SHU   

  1. School of Energy and Power Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China
  • Received:2020-04-17 Online:2021-02-05 Published:2021-02-09
  • Contact: Yingming XIE

摘要:

CO2水合物作为一种新型的蓄冷介质,具有良好的应用前景。在蓄冷系统中,CO2水合物的生成量对蓄冷量有直接影响,但CO2水合物生成计算较为复杂,进而导致系统蓄冷量的计算同样复杂起来,因此建立能够快速分析和预测系统中水合物生成量的模型具有实际意义。本文介绍了可以解决复杂问题的BP神经网络模型(BP)和灰色关联预测模型[GRM(1,n)],并利用Matlab语言编程建立了GRM(1,n)-BP神经网络组合模型来预测水合物的生成量。本文选取实验系统的数据利用3种模型分别进行了预测,并将3种模型的结果与实验结果进行对比,结果表明,GRM(1,n)-BP神经网络组合模型的准确性和稳定性效果更好。最后通过考察充注压力这一单一变量对水合物生成量的影响,并对比模型预测结果,进一步验证了GRM(1,n)-BP神经网络组合模型的准确性。

关键词: 二氧化碳, 水合物, 神经网络, 预测

Abstract:

As a new cold storage medium, CO2 hydrate has a good application prospect. In the cold storage system, the amount of CO2 hydrate generation has a direct impact on the amount of cold storage, but the calculation of the amount of CO2 hydrate generation is complicated, which leads to the calculation of the amount of cold storage in the system is also complicated. Therefore, it is of practical significance to establish a model that can quickly analyze and predict the amount of hydrate production in the system. In this paper, BP neural network model (BP) and grey relational prediction model [GRM(1,n)] which can solve complex problems were introduced, and GRM(1,n)-BP neural network combination model was established by Matlab programming language to predict hydrate production. Three models were selected to predict the data of the experimental system, and the results of the three models were compared with the experimental results. The results showed that the GRM(1,n)-BP neural network combination model has better accuracy and stability. Finally, the accuracy of the GRM(1,n)-BP neural network combination model by investigating the influence of the single variable of charging pressure on hydrate production and comparing the predicted results of the model was further verified.

Key words: carbon dioxide, hydrate, neural networks, prediction

中图分类号: 

版权所有 © 《化工进展》编辑部


本系统由北京玛格泰克科技发展有限公司设计开发 技术支持:support@magtech.com.cn