Control Of The Alumina Concentration Based On Data-Driven

The aluminum electrolysis is a highly energy-consuming and highly polluting process industry.With the application of computer technology and information technology in the aluminum electrolysis industry today,the development of aluminum electrolysis technology has become more information-based and intelligent.By understanding the production process and parameter data characteristics of the electrolytic cell,and analyzing the data during the operation of the electrolytic cell,it is known that the reasonable control of the aluminum oxide concentration in the aluminum electrolytic cell has an important role in improving the current efficiency and reducing the energy consumption of the aluminum electrolytic production process.In order to facilitate the effective prediction of the alumina concentration in the electrolytic cell,a data-driven neural network prediction algorithm is used to predict the alumina concentration in the electrolytic cell.Use expert knowledge to select 8 parameter data,such as electric current,operating voltage,feeding amount,aluminum tapping volume,aluminum level,resistance,molecular ratio and electrolyte level,which have a greater impact on the alumina concentration.Perform Lagrange interpolation processing on the missing values of the parameter data and perform box plot detection and noise reduction processing on the outliers,and then select the 8 processed parameter data as the input of the BP neural network alumina concentration prediction model,alumina concentration As the output data of the prediction model,a BP neural network prediction model for alumina concentration prediction was established.In order to improve the prediction accuracy and training speed of the BP neural network,the L-M optimization algorithm is used to optimize the neural network training.Using the data collected by industrial production as training data,it is divided into training set(70%),validation set(15%)and test set(15%).After calculation,the mean square error of BP neural network prediction model verification and testing are: 0.013581,0.013021.Through the inspection of industrial production data,the prediction accuracy of the model is 0.1,and a neural network prediction model that meets the requirements of industrial production is obtained.Finally,according to the alumina concentration predicted by the BP neural network prediction model,combined with the principle of alumina concentration change and production practice,it is proposed to adjust the alumina concentration value by adjusting the reference blanking interval of the electrolytic cell to ensure that the alumina concentration is in the desired range [ 1.5%,2.0%].