样本自选择回归分析算法在转炉炼钢中的应用

Application of the Regression Algorithm Based on Sample-Self-Selection in BOF Steelmaking

针对转炉炼钢静态控制模型广泛采用的常规回归分析算法采用目标炉次的前几炉次冶炼数据作为样本,模型预测准确度低的问题,提出了一种基于样本自选择的回归分析算法。该算法从实际生产数据中自动选取一定数量的合适样本来构建回归分析预测模型,实现吹氧量、冷却剂加入量、终点温度和终点碳含量的预测。通过某钢厂120t转炉Q235B钢种的实际生产数据与该算法、常规回归分析算法和BP神经网络算法进行预测结果比较,表明本算法具有预测准确度高,综合预测效果好等优点。

As a static control algorithm of BOF steelmaking,general regression algorithm chooses a few last smelting data as the sample without appropriate screening,and the accuracy of prediction was not high.To solve this problem,the regression algorithm based on sample-self-selection was proposed.The algorithm selected a certain number of appropriate samples from the actual production data automatically to build a regression model,implementing the predictions of oxygen consumption,coolant consumption,end-point temperature and end-point carbon content.Through comparing operational data and the prediction effects among general regression algorithm,BP neural network algorithm and the regression algorithm based on sample-self-selection for 120t converter,the results show that this algorithm has the advantage of high accuracy and effective comprehensive prediction.