摘要
针对高炉炼铁还原过程中非线性和大时滞等特点造成温度监测难度大的困境,提出一种融合数据分解、机器学习和误差修正的高炉铁水温度组合预测新模型.首先,引入带自适应白噪声的完备集合经验模态分解方法对铁水温度序列进行分解处理,通过提取不同频率的规律特征,使复杂的非线性序列转化为规律性较强的子序列;随后,采用相关向量机对子序列进行学习,充分挖掘铁水温度序列的信息,获得精度较高的预测结果;最后,将对铁水温度影响较大的硅含量和富氧率等相关因素作为辅助参数,使用经主成分分析处理后的辅助参数序列对预测结果进行修正,提高模型的预测准确性.结果表明:相较于整合移动平均自回归模型等传统模型,所提出的新模型综合性能更优,即平均绝对误差百分比减小53.57%,铁水温度为±10℃范围内的预测命中率提高25%.所提出的模型为实现高炉温度实时精细化调控提供了理论支撑,对保证炉况稳定、提升产品质量和降低冶炼能耗具有重大实际意义.
The reduction process of ironmaking is difficult to be controlled, which is caused by its characteristics of non-linearity and large time delay. Motivated by data decomposition, machine learning and error correction technologies,a novel hybrid prediction model is proposed for blast furnace hot metal temperature in this paper. Firstly, the complete ensemble empirical mode decomposition with adaptive noise is introduced to decompose the time series of hot metal temperature. The complicated non-linear time series are transformed into various sub-components by extracting the regular with different frequencies. Then, the relevance vector machine(RVM) is used to learn the rules of subsequences, and the information of the molten iron temperature sequence is fully mined to obtain a prediction result with high accuracy.Finally, the auxiliary parameter sequence processed by principal component analysis is used to modify the prediction results, improving the prediction accuracy of the model. The results show that compared with traditional models such as autoregressive integrated moving average model, the proposed model has better overall performance. The average absolute error percentage is reduced by 53.57%, and the predicted hit rate within the range of ±10℃ for the hot metal temperature is increased by 25%. The model has important practical significance for ensuring stable furnace conditions, improving product quality and reducing smelting energy consumption.
作者
廖亚楠
王业林
李萌
肖清泰
王华
LIAO Ya-nan;WANG Ye-lin;LI Meng;XIAO Qing-tai;WANG Hua(State Key Laboratory of Complex Nonferrous Metal Resources Clean Utilization,Kunming University of Science and Technology,Kunming Yunnan 650093,China;Faulty of Metallurgical and Energy Engineering,Kunming University of Science and Technology,Kunming Yunnan 650093,China;Department of Electrical and Computer Engineering,University of Central Florida,Orlando 32816,United States)
出处
《控制理论与应用》
EI
CAS
CSCD
北大核心
2022年第11期2177-2184,共8页
Control Theory & Applications
基金
云南省教育厅科学研究基金项目(2021J0063)
云南省科技厅科技计划项目(202101AU070031)
云南省基础研究计划项目(202101BG070127)资助。
关键词
机器学习
相关向量机
CEEMDAN
误差修正
铁水温度
预测
machine learning
relevance vector machine
CEEMDAN
error correction
hot metal temperature
prediction
