摘要
本文利用回声状态网络建立了转炉终点钢水碳含量和温度预测模型,并应用遗传算法优化其主要参数。选用某钢厂3座120t转炉的实测数据对该网络模型进行离线训练和仿真测试,结果表明回声状态网络模型比BP神经网络模型在预测精度上有所提高。这为开展转炉实时预测工作提供了方法指导。
An ESN model of endpoint temperature and the carbon content was set up,and the main parameters of the model were optimized by the genetic algorithm.The practical data of three 120t converters from a steel plant was selected for off-time training of the network model and simulating test.The results showed that the echo state network model was better than BP neural network model for improving the prediction accuracy.It provided method guidance for real-time forecasting.
出处
《冶金丛刊》
2011年第3期4-7,共4页
Metallurgical Collections
关键词
转炉终点预测
回声状态网络
遗传算法
prediction of the endpoint of BOF
echo state network
genetic algorithm
