欢迎访问过程工程学报, 今天是

过程工程学报 ›› 2025, Vol. 25 ›› Issue (6): 579-589.DOI: 10.12034/j.issn.1009-606X.224392

• 研究论文 • 上一篇    下一篇

基于MIC特征选择和WOA-LSSVM优化的阳极铜质量预测研究

熊文真1， 徐建新2， 熊英1*   

  1. 1. 信阳职业技术学院信息与通信工程学院，河南 信阳 464000 2. 昆明理工大学冶金与能源工程学院，云南 昆明 650093
  • 收稿日期:2024-12-23 修回日期:2025-04-18 出版日期:2025-06-28 发布日期:2025-07-01
  • 通讯作者: 熊英 461105526@qq.com
  • 基金资助:
    国家重点研发计划项目;科技厅重大专项课题

Research on anode copper quality prediction based on MIC feature selection and WOA-LSSVM optimization

Wenzhen XIONG1,  Jianxin XU2,  Ying XIONG1*   

  1. 1. School of Information and Communication Engineering, Xinyang Vocational and Technical College, Xinyang, Henan 464000, China 2. School of Metallurgical and Energy Engineering, Kunming University of Science and Technology, Kunming, Yunnan 650093, China
  • Received:2024-12-23 Revised:2025-04-18 Online:2025-06-28 Published:2025-07-01

PDF

摘要： 电解铜精炼过程中，阳极板中铜含量对电解效率至关重要。以混合铜精矿和粗铜等15种元素质量作为自变量，阳极板的铜元素质量作为因变量，利用最大信息系数(MIC)分析了54个具有代表性的测试数据集中各元素间的非线性相关性。结果表明，混合铜精矿的As含量和粗铜(外购)的Sb含量与阳极板铜含量的相关性最高，MIC值分别约为0.8228和0.8362。基于此，构建了鲸鱼算法优化的最小二乘支持向量机(WOA-LSSVM)回归预测模型，对阳极板铜元素质量进行预测。WOA-LSSVM模型具有较高预测精度，R2达0.9245，均方根误差(RMSE)较小，WOA-LSSVM组合模型对阳极板铜含量的预测精度比其他模型高出4.45%~123.05%。非线性分析方法能够有效捕捉阳极铜生产过程中不同因素之间的复杂关系，结合非线性分析方法和机器学习技术，可以提高阳极铜质量控制的实时性和适应性。

关键词: 阳极铜质量, 控制预测, 最大信息系数, WOA-LSSVM, 机器学习

Abstract: During the electrolytic copper refining process, the copper content in the anode plate is crucial to the electrolysis efficiency. Fifteen elemental qualities of mixed copper concentrate and crude copper are taken as independent variables, while the copper element quality of the anode plate is considered as the dependent variable. The maximal information coefficient (MIC) was used to analyze the nonlinear correlations among the elements in 54 representative test datasets. The study found that the arsenic (As) content in the mixed copper concentrate and the antimony (Sb) content in the purchased crude copper had the highest correlation with the copper content in the anode plate, with MIC values of approximately 0.8228 and 0.8362, respectively. Based on these findings, a whale optimization algorithm-optimized least squares support vector machine (WOA-LSSVM) regression prediction model was constructed to predict the copper element quality of the anode plate. Experimental results indicated that the WOA-LSSVM model has a high prediction accuracy, with an R2 value reaching 0.9245 and a low root mean square error (RMSE). The prediction accuracy of the WOA-LSSVM hybrid model for anode plate copper content was 4.45% to 123.05% higher than that of other models. Nonlinear analysis methods can effectively capture the complex relationships between different factors in the production process of anode copper. Combining nonlinear analysis methods with machine learning techniques can improve the timeliness and adaptability of anode copper quality control.

Key words: anode copper quality, control forecasting, maximum information coefficient, WOA-LSSVM, machine learning