Anode effect prediction based on Light Gradient Boosting Machine

The fault detection of the anode effect is of great significance in the process of aluminum electrolysis. Effective prediction of the anode effect can increase the output of the electrolytic cell and reduce the power consumption. In this paper, data is extracted from the real-time production process of aluminum electrolysis. In order to remove the redundancy and noise in the data and save the time cost of manual selection of features, the principal component analysis method is used to optimize other algorithms, and then compare with LightGBM. It has been proved by experiments that the LightGBM model not only improves the prediction accuracy of the anode effect to 99.9%, but also advances the fault prediction time to 40 minutes, which has high practical application value.