中文摘要：

在电解铝的实际生产过程中,由于不能实时监控槽电压的变化情况,容易发生电压摆动的问题。为了实时监控槽电压的变化以及预防电解槽的电压摆,提出了基于主成分分析（PCA）的极限学习机（ELM）多神经网络结构模型,用于铝电解生产过程槽电压预测。一方面,将极限学习机方法同主成分分析方法相结合,将高维输入变量压缩处理为低维主元变量,简化极限学习机模型,提高主成分分析极限学习机（PCA-ELM）算法的泛化性能。另一方面,将多个PCA-ELM子神经网络按照连接权值综合起来,建立铝电解生产过程槽电压的预测模型,进一步提高多神经网络模型的预测能力和预测精度。通过实际数据仿真结果表明,多神经网络预测模型能够准确的实时监控槽电压以及预防电压摆。

英文摘要：

In the actual production process of aluminum,can not real-time monitoring the change of cell voltage,voltage swing will be occurred in this process.In order to real-time monitoring the change of cell voltage and prevent cell voltage swing,a model of cell voltage of the aluminum electrolysis production process is proposed based on PCA-ELM multiple neural network structure.On the one hand,the method of extreme learning machine combined with principal component analysis method,to make the high dimensional input variable compression into low dimensional principal component variables.To simplified extreme learning machine model,and improve the generalization performance of the algorithm which the principal component analysis and extreme learning machine（PCA- ELM）.On the other hand,the multiple PCA- ELM sub neural network are combined according to the connection weights of neural network.To establish model of cell voltage of aluminum electrolytic production process,and improve the prediction of a neural network model and the prediction precision.Through the actual data simulation results show that neural network prediction model can more accurate real-time monitoring the cell voltage and to prevent voltage swing.