Abstract: A lot of useful information, which hide in the production data of aluminum electrolysis, are the important criterions to assess the situation of the cell. The influence of process parameters for cell states is analyzed in line with the principle of aluminum electrolysis process. According to the features of data of aluminum electrolysis, fuzzy clustering is used to analyze cell states. According to the problem of the noise in the aluminum electrolysis data, the Noise-Canceling FCM algorithm （NCFCM algorithm） is proposed so as to solve the problem which the traditional FCM algorithm is sensitive to outliers into a local optimum. NCFCM algorithm is used to cluster the actual production data of an aluminum factory. The results show that the accuracy rate of reduction cell clustering by NCFCM algorithm is more than 85% and clustering results can provide a reference for the production of aluminum electrolysis.

Key words: aluminum electrolysis, clustering analysis, Noise-Canceling Fuzzy C-Means（NCFCM） algorithm, classification of cell states

摘要： 铝电解生产数据中所存在的隐藏信息，是判断槽况状态的重要依据。根据铝电解的工艺原理，分析了工艺参数对铝电解槽况的影响;根据铝电解的数据特点，提出了利用模糊聚类的方法分析槽况。针对铝电解生产数据的存在噪声的问题，提出了具有除噪功能的FCM算法（NCFCM算法），以解决因传统FCM算法对孤立点敏感而陷入局部最优的问题;利用NCFCM算法对某铝厂的实际生产数据进行聚类分析。研究结果表明，利用NCFCM算法对槽况聚类的正确率达到90%以上，聚类结果可以为铝电解生产提供参考。

关键词: 铝电解, 聚类分析, 具有除噪功能的模糊C均值（NCFCM）算法, 槽况分类