Abstract: It is difficult to accurately describe probability distribution and to do correlation analysis of operating multiple parameters in aluminum reduction cells. A correlation analysis method based on mixed Copula model was proposed to resolve the problem. First of all, non-parametric kernel density estimation method was used to establish edge density function of variables in the case of unknown distribution type. Secondly, proportion of contribution of different Copula functions could be adjusted by weight parameters in the joint distribution function of multiple parameters. Finally, the maximum likelihood method was used to estimate the mixed model parameters. By using 1824 groups real data of 170 kA operating aluminum smelter from a factory, test results showed that this method was better than other single Copula function because three distance indicators were 0.3169, 0.6239 and 0.9276. It was an effective way under super-low-voltage condition to do multi-parameter correlation analysis with non-steady-state and non-uniform features.

Key words: correlation analysis, Copula, model, distributions, electrolysis

摘要： 针对铝电解大型预焙槽操作参数较多且彼此耦合性强，难以进行准确的概率分布描述和相关性分析问题，提出一种基于混合Copula模型的铝电解槽多参数相关性分析方法。在铝电解槽参数分布类型未知的情况下，首先利用非参数核密度函数估计建立变量的边缘密度函数；再构建基于混合Copula模型的多变量联合分布函数，并通过权重参数调节不同类型Copula函数的贡献比重；最后利用极大似然法对模型参数进行估计。对取自某厂170 kA铝电解槽的1824组真实样本数据进行实验，结果得到的3种距离指标分别是0.3169、0.6239和0.9276，均优于其他单一Copula函数，表明本方法是对超低电压下具有非稳态非均一特征的多参数进行相关性分析的一种有效途径。

关键词: 相关性分析, Copula, 模型, 分布, 电解