Anode effect forecast in aluminum electrolysis based on an extreme learning machine

A cell resistance forecast model is established based on an extreme learning machine (ELM) algorithm. Considering a harsh environment in aluminum electrolysis, there exist outliers in measured data. These outliers may affect the performance of an ordinary ELM algorithm, and even destroy the well-trained model. This paper introduces a modified extreme learning machine algorithm subject to the outliers. Based on the distributed measurement of current through the anode rod, the cell resistance can be calculated and the cell resistance at the next time point can be forecast. Then the cell resistance is made to contact with the anode effect and the alarm criterion is set. Real industrial data have been applied to verify the effectiveness and accuracy of this proposed scheme. © All right reserved.