第二届复杂系统与工程大会 (CSEC2021)

36 / 2021-08-17 15:15:41

State prediction of aluminum cell based on K-LSTM algorithm

aluminum electrolysis,cell state prediction,LSTM,time series

Aluminum electrolytic process is a very complex industrial process meanwhile lot of data produced in this production process. In order to facilitate the industrial personnel to grasp the change of the aluminum reduction cell in time, here we use the improved K-LSTM algorithm to predict the aluminum electrolytic cell state. The algorithm combines the characteristics of data changes, and proposes to solve the problem of sample imbalance in the LSTM forget gate unit, and eliminate the sample imbalance by setting the weight. The algorithm can effectively predict the state of the aluminum electrolytic cell, especially the ratio of the prediction of the sudden change period of the cell state to the model before the improvement can predict the change of the cell state more quickly, which  has a very important effect on the aluminum electrolysis industry with large time lag. It can predict the abnormality of the cell in advance, and experts can make the expected reduction in loss in time.