中国有色冶金

Anode effect is the most frequent fault in the aluminum electrolysis production, and accurate prediction of the anode effect can reduce energy consumption and reduce accidents. Starting from deep learning, this paper proposes a prediction model based on stacked noise reduction autoencoder and long short term memory network. The stacked noise reduction autoencoder is used to find key fault feature information, and meanwhile the long short term memory network is used to realize fault diagnosis. Finally, the historical production data of an aluminum factory is collected to verify the performance of the model. The experimental results show that the prediction accuracy and -F-1 score of the model are 9756% and 09686, respectively. This paper makes a comparative analysis of the BP neural network, generalized regression neural network, LSTM and SDAE-RF. The experimental results show that the SDAE-LSTM model constructed in this paper has the best performance with more accurate prediction of the anode effect, which has important guiding significance for the actual production of aluminum electrolysis.