面向铝工业基于PSO+LSTM的电解槽氧化铝浓度预测研究  

Prediction study of alumina concentration in electrolytic cell based on PSO+LSTM for aluminum industry

作　　者：朱家华 徐杨 曹斌 程建国 Zhu Jiahua;Xu Yang;Cao Bin;Cheng Jianguo(College of Big Data and Information Engineering,Guizhou University,Guiyang 550025,China;Guiyang Aluminum and Magnesium Design and Research Institute Co.,Ltd.,Guiyang 550009,China;Chinalco Intelligent Technology Development Co.,Ltd.,Hangzhou 311103,China)

机构地区：[1]贵州大学大数据与信息工程学院,贵州贵阳550025 [2]贵阳铝镁设计研究院有限公司,贵州贵阳550009 [3]中铝智能科技发展有限公司,浙江杭州311103

出　　处：《轻金属》2023年第7期21-27,共7页Light Metals

摘　　要：针对氧化铝浓度在高温、强腐蚀、强磁场的环境中无法实时在线测量及测量精度不高的问题,提出基于PSO(粒子群算法)优化LSTM(长短期记忆神经网络)的电解槽氧化铝浓度预测方法。首先,根据铝电解生产的相关知识,确定影响氧化铝浓度的相关参数;然后,结合基于PSO(粒子群算法)优化LSTM(长短期记忆神经网络)模型对氧化铝浓度进行预测;最后,通过某铝厂实际生产数据对模型进行验证。与现有方法相比,所提算法在原来只用阳极导杆分布电流和极间电压(阳极导杆到阴极钢棒之间的电压)两项参数的前提下,增加阳极导杆等距压降一项参数,进一步缩短了氧化铝浓度预测的时间,并且提高氧化铝浓度预测的准确度。Since the alumina concentration in high temperature,strong corrosion and strong magnetic field environment cannot be measured in real time and the measurement accuracy is not high,the LSTM(Long short-Term Memory)method based on PSO(Particle Swarm Optimization)for predicting alumina concentration in electrolytic cell was proposed.Firstly,based on aluminum electrolysis production knowledge,relevant parameters affecting alumina concentration are determined;Then,the LSTM model was optimized based on PSO to predict alumina concentration;Finally,the model was verified by the actual production data of one aluminum smelter.Compared to existing methods,the proposed algorithm is based on the premise that except for the distribution current of the anode rod and the voltage between anode rod and cathode steel bar,another parameter for equidistant voltage drop of anode rod is added,thus reducing the prediction time of alumina concentration and improving the accuracy of alumina concentration prediction.

关 键 词：氧化铝浓度 铝电解 LSTM PSO 浓度控制 

分 类 号：TF821[冶金工程—有色金属冶金]