摘要
电流效率是铝电解过程中最重要的经济指标,通过对电流效率的测量,可以有效管理出铝量,而目前电流效率的测量方法普遍具有周期较长、精确率低等缺点。针对电解槽在不同槽况下的电流效率不同的特征,提出了槽况分类的多支持向量机模型预测电流效率的方法,利用铝电解生产的历史数据作为样本,建立采用模糊聚类的槽况分类模型,得到三种槽况类别,对每个子类样本建立支持向量机预测子模型,经过子模型融合,得到电流效率最终预测模型。仿真结果表明,上述多分类预测模型命中率较高,可准确预测电流效率,用于指导实际电解铝的生产。
Current efficiency（CE） is the most important economic indicator in the process of aluminum electrolysis, the amount of aluminum can be effectively managed by predicting the current efficiency. The current method of measuring the CE generally has the shortcomings of longer period and low accurate rate. In view of the characteristics of different cell states exist in the process of aluminum electrolysis, we put forward a method of multi support vector machine （SVM） model based on the classification of cell states to predict the CE, and the historical data of aluminum electrolysis was used as a sample. First, we buih the model to classify the cell states based on Fuzzy clustering to get three types of cell states. Then, we used the SVM to build predict sub-model on the base of each subclass, so that we can get the CE prediction model based on the classification of cell states. This categorical predicting model has higher hit rate, can accurately predict the amount of aluminum and be used to guide actual production.
出处
《计算机仿真》
北大核心
2017年第1期288-291,387,共5页
Computer Simulation
基金
内蒙古自治区研究生科研创新基金资助项目(S20151012711)
国家自然基金资助项目(61164018)
关键词
铝电解
电流效率预测
槽况分类
模糊聚类
支持向量机
Aluminum electrolysis
Current efficiency prediction
Classification of cell states
Fuzzy clustering
Support vector machine（SVM）
