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

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.

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.