基于自适应遗忘因子极限学习机的高炉煤气预测  

Online Prediction Method for Generation and Consumption of Blast Furnace Gas Based on Adaptive Forgetting Factor Extreme Learning Machine

作　　者：孙雪莹[1,2] 胡静涛[1,2] 王卓[1,2] 张吉龙[1] 

机构地区：[1]中科院沈阳自动化研究所,沈阳110016 [2]中国科学院大学,北京100049

出　　处：《计算机测量与控制》2017年第7期235-238,共4页Computer Measurement &Control

基　　金：中国科学院重点部署项目(KGZD-EW-302);中国科学院科技服务网络计划(KTJ-SW-STS-159);辽宁省科学技术计划项目(2015020140)

摘　　要：高炉煤气是钢铁企业重要的二次能源,其产生量和消耗量的实时准确预测对高炉煤气系统的平衡调度具有重要作用;但由于高炉煤气系统工况多变、产消量数据波动较大,给高炉煤气产消量的准确预测带来了很大的挑战;为此,通过对煤气产消量数据特征的深入分析,提出了一种基于自适应遗忘因子极限学习机(AF-ELM)的在线预测算法;在序贯极限学习机的基础上,引入遗忘因子逐步遗忘旧样本,通过预测误差反馈机制,自适应的调节遗忘因子,从而提高预测方法对系统工况的动态变化的适应能力,提高预测精度;将该算法应用于钢铁企业的高炉煤气产消量在线预测,实验结果表明与序贯极限学习机相比,该预测方法在系统工况变化的情况下能保持较高的预测精度,更适合于高炉煤气产消量的在线预测。Blast furnace gas is an important byproduct in iron and steel plants, and prediction of its generation and consumption has a great effect on balance and scheduling of gas system. However, the accurate prediction poses a significant challenge because of the unstable conditions of the blast furnace gas system and the fluctuation of data. To solve this problem, an online prediction method based on adaptive forgetting factor extreme learning machine （AF-ELM） is proposed. Dynamic adaptability of online sequential extreme learning machine is improved by introducing forgetting factor to gradually forget of the old samples. And the forgetting factor is adaptively updated by prediction error, which improves the prediction accuracy. The case study on the online prediction in iron and steel plants shows that compared with online sequential extreme learning machine, the proposed method achieve higher prediction accuracy in changing conditions, and more suitable for online prediction of generation and consumption of blast furnace gas.

关 键 词：高炉煤气 在线预测 极限学习机 遗忘因子 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]