Prediction of calciner outlet temperature based on KPCA and BiLSTM

Journal Information

J. Hefei Univ. Tech. (Nat. Sci.), Feb.2023, Vol.46 No.2: 169-174

DOI: 10.3969/j.issn.1003-5060.2023.02.005

Authors

MENG Ren $ ^{1} $, DONG Xueping $ ^{1} $, GAN Min $ ^{2} $

(1. School of Electrical Engineering and Automation, Hefei University of Technology, Hefei 230009, China; 2. School of Mathematics and Computer Science, Fuzhou University, Fuzhou 350108, China)

Abstract and Keywords

Abstract: In the cement production process, the outlet temperature of the calciner is a very important process parameter. In order to cope with the diversity of outlet temperature variables, this paper proposes a combined model of temperature prediction using kernel principal component analysis (KPCA) and bidirectional long short-term memory (BiLSTM) neural network to predict the outlet temperature of the calciner. Through the KPCA, the main components that affect the temperature variables at the outlet of the calciner are selected to achieve the data dimensionality reduction. The reduced main components are used as the input of BiLSTM, and the temperature at the outlet of the calciner is used as the output of BiLSTM. After BiLSTM neural network training, the outlet temperature prediction model of the calciner is obtained. The comparison and verification show that the temperature prediction model with KPCA-BiLSTM has better prediction accuracy.

Keywords: cement calciner; outlet temperature; kernel principal component analysis(KPCA); bidirectional long short-term memory(BiLSTM) neural network; dimensionality reduction; prediction

作者信息

孟忍 $ ^{1} $，董学平 $ ^{1} $，甘敏 $ ^{2} $

(1. 合肥工业大学电气与自动化工程学院, 安徽合肥 230009; 2. 福州大学数学与计算机科学学院, 福建福州 350108)

摘要和关键词

摘要: 水泥生产过程中,分解炉出口温度是非常重要的工艺参数,为了应对出口温度变量的多样性,文章提出一种核主成分分析(kernel principal component analysis,KPCA)与双向长短期记忆(bidirectional long short-term memory,BiLSTM)神经网络相结合的温度预测组合模型用来预测分解炉的出口温度。通过KPCA筛选出影响因素的主成分从而达到数据降维目的,将降维后的主成分作为BiLSTM神经网络的输入,分解炉出口温度作为BiLSTM神经网络的输出。经BiLSTM神经网络训练,得到分解炉出口温度预测模型。通过对比验证表明,使用KPCA-BiLSTM相结合的温度预测模型具有较好的预测精度。

关键词: 水泥分解炉；出口温度；核主成分分析(KPCA)；双向长短期记忆(BiLSTM)神经网络；降维；预测

Funding

国家自然科学基金资助项目（61673155）

合肥工业大学学报(自然科学版)

Navigation