基于CEEMDAN-SVM-LSTM的高炉煤气利用率组合预测  

Combined forecast of blast furnace gas utilization rate based on CEEMDAN-SVM-LSTM

作　　者：石琳[1] 刘文慧 曹富军 王晋建 SHI Lin;LIU Wenhui;CAO Fujun;WANG Jinjian(School of Science,Inner Mongolia University of Science and Technology,Baotou 014010,China;School of Information and Engineering,Inner Mongolia University of Science and Technology,Baotou 014010,China)

机构地区：[1]内蒙古科技大学理学院,内蒙古包头014010 [2]内蒙古科技大学信息与工程学院,内蒙古包头014010

出　　处：《中国测试》2023年第1期86-91,共6页China Measurement & Test

基　　金：国家自然科学基金资助项目(11801287);内蒙古自然科学基金资助项目(2020MS06010);内蒙古自治区青年科技英才支持计划项目(NJYT20B15);鹿城英才高校专业技术人才基金项目(0701012003)。

摘　　要：煤气利用率是高炉炉况稳定和耗能的重要指标之一。为提高煤气利用率的预测精度,提出一种基于CEEMDAN-SVM-LSTM的组合模型对其进行预测。首先利用CEEMDAN(自适应噪声完备集合经验模态分解)将煤气利用率时间序列分解成6个模态量和一个趋势分量,对煤气流利用率的发展进行解耦;然后用LSTM(长短时间记忆人工神经网络)和SVM(支持向量机)分别对分解的高频模态和低频模态进行预测,最后将模型组合建立原始煤气利用率的组合预测模型。结果表明该组合模型的MAE(平均绝对误差)、MAPE(平均绝对百分比误差)、RMSE(均方根误差)和MSE(均方误差)分别为0.14、3.5%、0.18、0.032。与单一的SVM模型和LSTM预测模型对比,组合模型的精度更高。The gas utilization rate is one of the important parameters for the stability and energy consumption of the blast furnace.In order to improve the accurate prediction of gas utilization rate,a combined model based on CEEMDAN-SVM-LSTM is proposed to predict it.First,CEEMDAN(complete ensemble empirical mode decomposition with adaptive noise) algorithm is used to decompose the gas utilization rate time series into 6modal components and a trend component,which decouples the development of the gas flow utilization rate.Additionally,LSTM(long short-term memory) and SVM(support vector machine) are predicted to the decomposed high-frequency mode and low-frequency mode respectively,and finally combine the models to establish a complex prediction model of the original gas utilization rate.The results showed that the MAE(mean absolute error),MAPE(mean absolute percentage error),RMSE(root mean square error),and MSE(mean square error) of the combined model are 0.14,3.5%,0.18 and 0.032,respectively.Compared with the single SVM and the LSTM prediction model,the combined model has higher accuracy.

关 键 词：煤气利用率预测 快速傅里叶变换 CEEMDAN LSTM SVM 

分 类 号：TP3-05[自动化与计算机技术—计算机科学与技术]