高炉铁水硅含量预测评价与模型选择

Evaluation and Model Selection of Forecasting Silicon Content in Hot Metal of Blast Furnace

分析了文献中高炉铁水硅含量预测数据,结果表明:评价指标选择决定预测效果的评价结论和模型选择;命中率判据不是评价预测效果的最佳选择;相关系数及其检验是评价预测结果优劣的客观判据;相关系数及其检验、人工预测、随机游走模型可以作为以科学性、实用性和简便性为原则的预测评价判据基准。

Data in previous publications of forecasting silicon content in hot metal of blast furnace are analyzed. The result shows that evaluation of forecasting results depends on the used criteria. Hit ratio is not the most scientific criteria for forecasting evaluation in comparison with correlation coefficient between the measured and forecasted silicon content. The correlation coefficient and its test, judgment of blast furnace operator and random walk model should be considered as benchmark of the evaluation based on rational, practical and simple principle. Normalizing the scientific criteria of forecasting silicon content in hot metal is suggested.