基于智能模型的高炉煤气利用率预测

Gas utilization ratio prediction of blast furnace based on intelligent model

煤气利用率是影响高炉能量利用的关键参数,对高炉的稳定顺行有重要影响。在进行高炉生产参数时间间隔分析后,利用多种预测模型(自适应增强学习(Adaptive Boosting,AB)、随机森林(Random Forest,RF)、神经网络(Neural Network,NN))完成对煤气利用率的预测。首先,对高炉生产的历史数据进行收集整理;其次,在考虑高炉生产异常状态(如检修休风)下对特征参数完成预处理工作;最后,利用高炉参数间的时滞关系对煤气利用率建立预测模型,通过对预测模型的调参优化及评价,确定适合煤气利用率的预测模型,实现煤气利用率的提前预测,同时利用相同模型参数对原始数据集和预处理数据集进行预测并比较结果。结果表明,数据预处理对于预测结果的影响巨大,随机森林模型对煤气利用率的预测效果更好,可以用来指导高炉的实际生产。

Gas utilization ratio is a key parameter that affects the energy utilization of blast furnace,and it has an important relationship with the stable running of blast furnace.After time interval analysis of production parameters of blast furnace,several simple prediction models(adaptive boosting,random forest,neural network)are used to predict gas utilization rate.Firstly,the historical data of blast furnace production is collected and sorted.Secondly,the pretreatment of characteristic parameters is carried out considering the abnormal status of blast furnace production(such as overhaul and wind rest).Finally,using time-delay relationship between parameters of blast furnace,gas utilization ratio prediction model is set up.By parameter adjustion and optimization and evaluation of prediction model,gas utilization ratio prediction model is determined,to realize gas utilization ratio prediction in advance.At the same time,the same model parameters are used to predict the original data set and the pretreatment data set,and the results are compared.The results show that the data pretreatment has a great influence on prediction results,and the random forest model has a better prediction effect on gas utilization rate,which can be used to guide the actual production of blast furnace.