| 引用本文: | 蒋朝辉,李晞月,桂卫华,谢永芳,阳春华.分段线性回归和动态加权神经网络融合的高炉料位预测[J].控制理论与应用,2015,32(6):801~809.[点击复制] |
| JIANG Zhao-hui,LI Xi-yue,GUI Wei-hua,XIE Yong-fang,YANG Chun-hua.Blast furnace stockline prediction by segmented linear-regression and dynamic weighting neural network[J].Control Theory and Technology,2015,32(6):801~809.[点击复制] |
|
| 分段线性回归和动态加权神经网络融合的高炉料位预测 |
| Blast furnace stockline prediction by segmented linear-regression and dynamic weighting neural network |
| 摘要点击 2496 全文点击 1106 投稿时间：2015-01-12 修订日期：2015-03-18 |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| DOI编号 10.7641/CTA.2015.50033 |
| 2015,32(6):801-809 |
| 中文关键词 高炉 料位 预测 分段线性回归 动态加权 神经网络 |
| 英文关键词 blast furnace stockline prediction segmented linear-regression dynamic weighting neural network |
| 基金项目 国家自然科学基金重大项目(61290325), 国家自然科学基金创新研究群体科学基金项目(61321003)资助. |
|
| 中文摘要 |
| 针对高炉料位难以连续高精度测量的问题, 提出了一种基于分段线性回归和动态加权神经网络的高炉料位信息预测方法. 首先, 通过分析高炉布料机制和料位检测数据特点, 提出了一种面向雷达和机械探尺检测数据时间序列的联合划分方法, 用于提取高炉料位的周期性变化特征; 然后, 利用该变化特征构建分段线性回归模型, 获得能准确描述料位变化的回归曲线; 最后, 以回归统计指标为权重调节系数, 利用动态加权径向基神经网络对料位信息进行预测. 实例验证表明, 该方法融合了机械探尺检测数据精度高以及雷达检测数据连续性好的特点, 实现了高炉料位信息的实时有效预测. |
| 英文摘要 |
| Because of difficulties in continuously measuring the blast furnace stockline precisely, we propose a new prediction method based on segmented linear-regression (SLR) and dynamic weighting neural network (DWNN). According to blast furnace burden distribution schemes and data characteristics of blast furnace stockline, we design a combined division method based on the time series of radar and the mechanical stock rod data to extract periodical variation features of the blast furnace stockline. Then, a segmented linear-regression model is built based on the periodical variation features which help to obtain regression curves that precisely reflect the change of stockline. Finally, we take regression statistical indexes as coefficients of the regulation weights, and construct a dynamic weighting radial basis function (RBF) neural network model to predict the information of the blast furnace stockline. Case study indicates that the proposed method combines the high-precision of mechanical stock rod data with the continuity of radar data, and provides with real-time prediction for blast furnace stockline information effectively. |
|
|
|
|
|