GA-BP的铁钢界面铁水温度预测

铁钢界面铁水温度对炼钢生产的控制与优化具有重要意义.因此,为了更加准确地获取铁钢界面铁水温度,本文采用较大样本构建了基于遗传算法(GA)优化的BP神经网络铁水温度预测模型.对影响铁钢界面铁水温度的因素分析,选取了出铁时间、预处理时间、重罐时间、空罐时间、出铁铁水温度、预处理后铁水温度、铁水质量7个关键因素作为模型的输入,铁钢界面铁水温度作为输出.结果表明:GA-BP神经网络预测模型具有较好的性能表现,绝对误差10℃范围内的预测命中率达到了89%,平均绝对百分比误差(MAPE)为0.39%,均方根误差(RMSE)为6.41℃.相比于BP神经网络预测铁钢界面铁水温度,GA-BP模型具有更高的精确度和稳定性.

Prediction on temperature of hot metal on ironmaking and steelmaking interface based on GA-BP

The hot metal temperature of the ironmaking and steelmaking interface is significant to the control and optimization of steelmaking production. Therefore, in order to obtain the hot metal temperature more accurately, this paper establishes a prediction model on hot metal temperature with large samples based on BP neural network with optimized genetic algorithm(GA). On account of the analysis of influence factors of the hot metal temperature on the ironmaking and steelmaking interface, seven key factors including tapping time, pretreatment time, heavy tank time, empty tank time, hot metal temperature of tapping, hot metal temperature after pretreatment and hot metal quality are taken as input of the prediction model. Meanwhile, the hot metal temperature of the ironmaking and steelmaking interface is taken as output of the model. The results show that the GA-BP neural network prediction model has relatively better performance. The prediction hit rate of absolute error within 10 ℃ reaches 89%, MAPE is 0.39 %, and RMSE is 6.41 ℃. Compared with BP neural network, GA-BP model has higher accuracy and stability in predicting the hot metal temperature on the ironmaking and steelmaking interface.