【作者】 郑忠； 范计鹏； 蒋胜龙；

【Author】 Zheng Zhong;Fan Jipeng;Jiang Shenglong;College of Materials Science and Engineering, Chongqing University;

【机构】 重庆大学材料科学与工程学院；

【摘要】 氧气是钢铁企业生产所需的重要能源介质之一,转炉炉次耗氧量的准确预测有助于促进氧气能源网络供需平衡的精准调度,减少氧气放散。论文针对转炉炼钢过程氧气需求按冶炼炉次具有间歇性、影响因素复杂等特点,建立了一种基于机理分析与数据驱动相结合的转炉炉次耗氧量预测的LMBP神经网络模型。通过转炉耗氧量的机理分析,进行转炉耗氧量预测模型影响因素的提取与计算描述;以结合梯度下降法和高斯牛顿法优点的LMBP神经网络方法为基础建立数据驱动的耗氧量预测模型;利用Pearson相关性分析方法进行属性约简,设计隐藏层宽度及深度对比实验进行网络拓扑结构优化。利用某钢厂实例数据进行的仿真测试结果表明所建立的模型的转炉炉次氧气消耗量预测精度最高可达84. 80%,为进一步的现实生产氧气能源系统的精准调度提供参考依据。更多还原

【Abstract】 Oxygen is an important energy for iron and steel enterprises. Accurate prediction for oxygen consumption in the converter contribute to balance between supply and demand in oxygen energy network, and reduce oxygen emission. In process of converter steelmaking, oxygen demand has two important characteristics including the intermittent consumption caused by discrete charges and being influenced by many complex factors. To overcome these challenges, this paper proposes a LMBP neural network prediction model for oxygen consumption in the converter combined with mechanism analysis and data driving. Through the mechanism analysis of oxygen consumption in the converter, this paper extract and describe influencing factors of oxygen consumption prediction model; construct a data-driven prediction model based on the LMBP neural network which combines the advantages of the gradient descent method and the Gauss-Newton method; reduce the input attributes by Pears on correlation analysis method;optimize the network structure by carrying out experiment on the hidden layer in width and in depth.Finaliy, the simulation tests are carried out on the instance data extract from a steel plant. The experimental results show that the accuracy of the prediction model is up to 84.80%, which provides an avenue for precise scheduling in the realistic oxygen energy system.更多还原