Research On Support Vector Regression Of Soft-Sensor Measurement Model For Aluminum Reduction Cell Anode-Cathode-Distance

The cell voltage of the aluminum reduction cell is the important target of the energy consumption of the aluminum reduction cell, it is restricted to anode-cathode-distance(ACD), reduced the ACD is equal to reduces the energy consumption, therefore reduced the ACD is the key of the aluminum reduction cell energy consumption. The traditional energy conservation method of the energy conservation is realized through establish the cell voltage fluctuation model, but the energy conservation is realized few through the establishment ACD model to reduces the distance of ACD. Therefore, through analysis the influencing factor to establish the soft sensor model of the ACD is reduced the distance between electrodes, it has important meaning to reduce the aluminum reduction cell energy consumption.This article in the introduction of the process of the aluminum reduction and the energy conservation of the aluminum reduction, choosing affect the biggest effect of the aluminum reduction cell ACD is taken the object of study. After having analyzed in influence aluminum reduction process of the change factor of the aluminum reduction cell ACD, choosing the stem equal-space voltage, the aluminum reduction cell voltage, the anode resistance, the stem temperature and so on, studying the support vector machines(SVM) of the ACD soft-sensor measurement models, is realized the online estimate of the ACD. In view of the ACD's complexity, is proposed the support vector regression (SVR) algorithm, carrying on the model of ACD with the overlapping parameter optimization. Afterward aiming at the shortcoming which theε-SVR parameter dependence experience chooses and limiting condition, further is proposed geneticε-SVR, and has established the ACD soft sensor model.The experimental result showed that based onε-SVR aluminum reduction cell soft sensor model as a result of its algorithm's constraints and the choice of parameter is difficult, is caused the computing time to be big, time-consuming long and the shortcoming of parameter dependence experience chooses. But geneticε-SVR is saved the parameter selection time, strengthened the model real-time, and it is helped to the aluminum electrolysis process aluminum reduction cell ACD real-time online examination.