Application Of Data Mining In Comprehensive Analysis Of Cell Condition In Aluminum Electrolysis

Aluminum electrolysis production is a non-linear,complex variable and variable reaction time-delay process.In this process,the factors affecting the production status of aluminum electrolysis cells are difficult to control.At present,research on the cell condition is mostly based on simple mathematical statistics or modeling,and lacks sufficient application of historical data.In particular,production personnel rely on manual experience in actual production,and often cause significant losses when errors occur.Therefore,combining the data mining technology with the historical data of aluminum electrolysis production is of great significance for the analysis of cell condition in aluminum electrolysis.This paper first discusses the research status of the previous research on aluminum electrolysis production and the development trend of data mining technology,then uses the relevant data mining technology to deal with the missing values and outliers in the historical data,and analyze the data overview.Subsequently,by analyzing the meaning of each parameter produced in the production of aluminum electrolysis,nine different cell conditions were defined according to the numerical values of the aluminum level and the molecular ratio.Secondly,in order to reduce errors and improve speed of prediction,this paper uses the feature selection algorithm based on random forest to select the two parameters of aluminum level and molecular ratio.The results show that The training and prediction errors of the models used in this paper are fully in line with the production requirements of the aluminum plant.Among them,the root mean square error of the aluminum level training data is RMSE=0.4175,the R2 fitting result is 0.8241;the root mean square error of the molecular ratio training data is RMSE=0.0327,and the R2 fitting result is 0.8401;the root mean square of the aluminum level prediction data The error RMSE=0.44]3,the R2 fitting result is 0.7593;the root mean square error of the molecular ratio prediction data is RMSE=0.0350,and the R2 fitting result is 0.7807.Then,the cell condition analysis shows that the classification precision and recall rate of the groove data in the prediction data are 100%and 92.31%.Finally,by using the Python,PyQt5 and Qt Designer tools are combined to design and implement a data mining system based on analysis of cell condition in aluminum electrolysis.The system realizes functions such as data processing,data analysis,parameter prediction and visualization,which facilitates the production personnel to predict the values of aluminum level and molecular ratio parameters in advance in actual production,so as to master the cell condition and effectively guide aluminum electrolysis production.