基于数据特征的加热炉钢温预报模型

针对加热炉工业过程具有复杂、非线性、时滞性的特点和钢坯出炉温度预报问题，提出了一种基于数据特征的改进主元回归(PCR)加热炉钢温预报模型的建立方法.首先通过对原始数据进行同步化处理来解决各数据变量间存在的时间滞后问题;然后提取生产过程中各批次钢坯的统计特征和熵特征，并依据一定顺序将这些特征排列组合，构造等长的数据特征向量;最后通过PCR方法建立过程变量的数据特征和钢坯出炉温度之间的回归预报模型.本文以某钢厂加热炉工业过程为背景进行实验仿真，采用实际生产数据求取建模参数，并对钢坯出炉温度预报进行了测试.实验的校验与误差分析表明，该方法在预测钢坯出炉温度方面具有更好的性能，且预测误差满足工业应用的精度要求.

Billet Temperature Predication Model of Reheating Furnace Based on Data Features

Considering that the industrial process of reheating furnace is with the characteristics of complexity， non-linearity and time delay， and the prediction of billet temperature is difficult to achieve， an improved principal component regression(PCR)and prediction method based on data features is proposed in this paper. The time-delay among different variables is solved at first by synchronization of the original data. Then statistic and entropy features are extracted from each billet of reheating furnace and these data features consist of a data feature vector orderly. Lastly， the prediction model between billet outlet temperature and data features of process variable is established by PCR. The proposed method is applied in the reheating furnace of a real steel factory， and the model parameter is reckoned based on actual operational data. The experiment results and error analysis indicate that this model is able to predict the billet steel outlet temperature， and the prediction error can satisfy the demands of industrial application.