文献类型：期刊文献

英文题名：Hybrid model of a cement rotary kiln using an improved attention-based recurrent neural network

作者：Zheng, Jinquan[1];Zhao, Liang[1,2];Du, Wenli[1,2,3]

机构：[1]East China Univ Sci & Technol, Key Lab Smart Mfg Energy Chem Proc, Minist Educ, Shanghai 200237, Peoples R China;[2]Tongji Univ, Shanghai Inst Intelligent Sci & Technol, Shanghai 200092, Peoples R China;[3]Box 293,Meilong Rd 130, Shanghai 200237, Peoples R China

年份：2022

卷号：129

起止页码：631

外文期刊名：ISA TRANSACTIONS

收录：;EI(收录号：20220911739835);WOS:【SCI-EXPANDED(收录号:WOS:000876692400005)】；

基金：This work was supported by National Natural Science Foundation of China (Basic Science Center Program: 61988101), the National Science Fund for Distinguished Young Scholars, China (61725301, 61925305), the Programme of Introducing Talents of Discipline to Universities (the 111 Project), China under Grant B17017 and the Fundamental Research Funds for the Central Universities, China.

语种：英文

外文关键词：Hybrid model; Rotary kiln; Attention mechanism; Recurrent neural network

摘要：A rotary kiln is core equipment in cement calcination. Significant time delay, time-varying, and nonlinear characteristics cause challenges in the advance process control and operational optimization of the rotary kiln. However, the traditional mechanism model with many assumptions cannot accurately represent the dynamic kiln process because kinetic parameters are difficult to obtain. This paper proposes a novel hybrid strategy to develop a dynamic model of a rotary kiln by combining a process mechanism and a recurrent neural network to address this issue. A time delay mechanism is used to estimate the kiln's residence time to compensate for the time delay. A long short-term memory model that combines an attention mechanism and an ordinary differential equation solver is proposed to capture the time-varying and nonlinear behaviors of the kiln process. Case studies from two real-world cement plants with different processing loads are used to verify the effectiveness of the proposed hybrid modeling strategy. The results show that the proposed method has better accuracy and robustness than the traditional methods. The sensitivity analysis of the proposed model also makes it practical for t control system design and real-time optimization.(c) 2022 ISA. Published by Elsevier Ltd. All rights reserved.