基于遗传算法优化BP神经网络的高炉喷煤优化

Optimization of Pulverized Coal Injection in Blast Furnace Based on BP Neural Network Optimized By Genetic Algorithm

摘要:

高炉炼铁是一个复杂的多变量系统,而现行的操作制度是基于炉长经验的参数设置模式,导致能源尤其是煤粉的消耗常常处于“盲目”状态；文章综合炼铁工艺理论和高炉专家经验,针对白云鄂博矿石冶炼的特殊性,采用筛选出的优化数据,利用遗传算法所固有的全局搜索性能优化BP神经网络模型的权值和阈值,分别建立了基于遗传算法优化BP神经网络的高炉喷煤量优化预测模型以及工艺指标(铁水含量及入炉焦比)预测模型；优化数据的利用使得上述模型可以根据高炉当前炉况输出喷煤量的最佳优化设定值,并预测出相对应的工艺指标变化趋势；实际应用表明,本方法能够给现场操作人员提供操作指导,实现高炉稳定顺行、提高经济效益的目的。

Abstract:

Blast Furnace iron making process is a complex multi-variable system, due to the operational parameters are set by BF masters with their experience, thus making the consumption of resources are often in a "blind" state. According to the analysis of mechanism of blast furnace and experiences of the blast furnace experts, considering the speciality of ore smelting in Bayan Obo, the coal injection quantity prediction model and process indicators ( content in molten iron and coke ratio) prediction model based on the BP neural network which weights and thresholds has been optimized through the genetic algorithm were established by using optimizing data. So the optimization setting value of coal injection quantity and the changing trend of process indicators can be given according these models. The results of practical application showed that, this method can provide decision-making guidance for operators and achieve the goals of stable operation and increase benefit in the process of BF iron making