改进灰色关联熵结合BP网络铁水脱硫率预测模型

A model for predicting desulfurization rate of molten iron based on combination of improved gray correlational entropy with neural network

提出改进灰色关联熵结合神经网络的铁水脱硫率预测模型,用以降低脱硫成本,提高脱硫效率。将灰色关联与改进熵权法结合,优化输入,选取铁水进站质量、进站温度、温降、喷吹时间、石灰消耗量、镁消耗量与进站硫量作为神经网络的输入量,构建预测模型,并与其他模型的预测结果进行了对比分析。结果表明,改进灰色关联熵结合神经网络模型的预测精度更高,与单一BP神经网络、深度神经网络、随机森林模型相比,平均误差分别降低了4.20%、3.83%、4.65%。

A prediction model for desulfurization rate of molten iron is proposed based on combination of improved gray correlational entropy with neural network to reduce the cost and improve the efficiency of desulfurization.The input is optimized by combining the gray correlation with the improved entropy weight method.The weight of intake molten iron,the intake temperature,the temperature drop,the injection time,the lime consumption,the magnesium consumption,and the initial sulfur content are selected as the inputs of the neural network,and the prediction model is constructed.The results are compared with those of other models.The results show that the new model exhibit the highest prediction accuracy.Compared with single BP neural network,deep neural network,and random forest model,the average error of the new model is reduced by4.20%,3.83%,and 4.65%,respectively.