摘要
针对硅锰合金埋弧熔炼过程的特点，提出了一种基于自适应递推最小二乘支持向量机（ARLSSVM）的合金成分在线预测模型。该模型以实测工况参数为数据集，当新增一个样本时，分别采用增长记忆递推算法、限定记忆递推算法和缩减记忆递推算法训练最小二乘支持向量机（LSSVM），有效避免高维矩阵的求逆，加快模型更新的速度。然后通过自适应模型匹配算子选择相应的预测输出模型，提高模型的预测精度。将此模型应用于30MVA硅锰合金埋弧炉冶炼过程合金成分在线预测，实际生产运行数据验证了此方法的有效性。
According to the characteristics of the silicon-manganese smelting process in a submerged arc furnace, an on-line model for prediction of silicon-manganese composition based on an adaptive recursive least square support vector machine （ARLSSVM） is proposed in the paper. In this model, three recursive algorithms （increased recuisive algorithm, fixed algorithm and decreased algorithm） are employed to train the least square support vector machines using the measured data, which avoids the difficulties in solving high-dimensional inverse matrix and makes the model be updated rapidly. Then the adaptive model selection operator is used to select the proper one as the prediction model according to the process conditions to improve the prediction precision. The method was applied to the online prediction of the silicon-manganese composition in a 30MVA submerged arc furnace smelting process, and the result showed its effectiveness.
出处
《高技术通讯》
CAS
CSCD
北大核心
2011年第11期1213-1218,共6页
Chinese High Technology Letters
基金
863计划（2009AA042124）和国家自然科学基金（60634020,608740697,60843002）资助项目.
关键词
硅锰合金
成分
在线预测
递推最小二乘支持向量机
自适应模型选择算子
silicon-manganese alloy, composition, on-line prediction, recursive least squares support vector machine, adaptive model selection operator
