引用本文: | 蒋朝辉,尹菊萍,桂卫华,阳春华.基于复合差分进化算法与极限学习机的高炉铁水硅含量预报[J].控制理论与应用,2016,33(8):1089~1095.[点击复制] |
JIANG Zhao-hui,YIN Ju-ping,GUI Wei-hua,YANG Chun-hua.Prediction for blast furnace silicon content in hot metal based on composite differential evolution algorithm and extreme learning machine[J].Control Theory and Technology,2016,33(8):1089~1095.[点击复制] |
|
基于复合差分进化算法与极限学习机的高炉铁水硅含量预报 |
Prediction for blast furnace silicon content in hot metal based on composite differential evolution algorithm and extreme learning machine |
摘要点击 3190 全文点击 2086 投稿时间:2015-08-24 修订日期:2016-06-01 |
查看全文 查看/发表评论 下载PDF阅读器 |
DOI编号 10.7641/CTA.2016.50696 |
2016,33(8):1089-1095 |
中文关键词 铁水硅含量 预报模型 复合差分 极限学习机 |
英文关键词 silicon content in hot metal prediction model composite differential extreme learning machine |
基金项目 国家自然科学基金重大项目(61290325), 国家自然科学基金创新研究群体科学基金项目(61321003)资助. |
|
中文摘要 |
针对铁水硅含量无法直接在线检测的问题, 本文提出了一种基于优化极限学习机(ELM)的高炉铁水硅含量
预报方法. 该方法利用复合差分进化算法(CoDE)的快速定位全局最优解的能力来优化极限学习机的输入权值和隐
层节点阈值, 在此基础上建立了基于复合差分进化算法优化极限学习机(CoDE–ELM)的高炉铁水硅含量预报模型.
以某钢铁厂2650 m3的高炉为例, 利用实际采集数据进行模型检验, 结果表明, 当绝对误差小于0:1时, 铁水硅含量的
预报命中率为89%, 均方根误差为0:047, 实际目标值序列与预报值序列的相关系数为0:851. 所建模型的预报结果
优于支持向量机(SVM)、前馈神经网络(BP–NN)、极限学习机以及差分优化极限学习机(DE–ELM), 对高炉炉温的
实际调控具有较好的指导意义. |
英文摘要 |
Considering the silicon content of the hot metal cannot be directly detected online, a prediction method
for silicon content in hot metal based on the optimized extreme learning machine (ELM) is proposed. The weights of
inputs and thresholds of hidden nodes in the extreme learning machine are optimized by a composite differential evolution
algorithm (CoDE) because of its ability of quickly locating the global optimum solution. With the optimized results, a
prediction method based on the composite differential evolution extreme learning machine (CoDE–ELM) is established.
The proposed method is verified by using the actual data on a 2650 m3 blast furnace of a steel plant. The verification
results show that when the relative prediction error is confined to 0.1, the hit rate is 89%, the root mean square error of
prediction is 0.047, and the correlation coefficient of the sequence of the actual target value and the predicted target value
is 0.851. Through experiments, it can be seen that the prediction results of the established model are much better than
that of the support vector machine (SVM), feedforward neural network, extreme learning machine (ELM) and differential
evolution optimized extreme learning machine (DE–ELM). Moreover, the model provides important guiding significance
to the temperature control of the blast furnace. |
|
|
|
|
|