改进深度信念网络的转炉耗氧量预测

Prediction of converter oxygen consumption in improved deep belief network

为降低炼钢能耗,提高炼钢的产量、质量及炼钢过程中的耗氧量预测精度,针对某钢厂的转炉,基于海量历史数据,提出一种基于深度学习的改进深度信念网络(DBN)转炉耗氧量预测模型。通过引入高斯伯努利受限玻尔兹曼机(GBRBM),解决传统DBN中受限玻尔兹曼机(RBM)所引起的在连续输入时造成的信息丢失问题。首先经过数据预处理,再采用灰色关联度法,找出影响耗氧量的主导因素,最后将其作为GBRBM-DBN模型的输入,建立GBRBM-DBN模型,并通过仿真验证该方案的可行性。结果表明:该方案能够准确地预测炼钢过程中的耗氧量,预测精度高,泛化性强,可为实际生产提供理论指导。

In order to reduce the energy consumption of steelmaking,improve the production and quality of steelmaking and the prediction accuracy of oxygen consumption in steelmaking process,an improved deep belief network model based on deep learning is proposed for the converter of a steel mill based on massive historical data.By introducing the Gauss Bernoulli restricted Boltzmann machine(GBRBM),the problem of information loss caused by continuous input of RBM in the traditional deep belief network is solved.Firstly,after data preprocessing,the grey correlation method is used to find out the dominant factors affecting oxygen consumption.Finally,it is used as the input of GBRBM-DBN model to establish GBRBM-DBN model.The feasibility of the scheme is verified by simulation.The results show that the scheme can accurately predict the oxygen consumption in the steelmaking process,with high prediction accuracy and strong generalization,which can provide theoretical guidance for actual production.