Research On Data-driven Optimization Strategy For Rare Earth Electrolysis Current Efficiency

Rare earth production process usually uses molten salt electrolysis to prepare rare earth monomers and alloys,where current efficiency is the key index in the process of molten salt electrolysis.In this thesis,we propose a data-driven current efficiency optimization strategy for rare earth electrolysis to improve current efficiency and reduce energy consumption by accurately predicting the electrolysis efficiency and predicting the electrolysis status and adjusting the control technology parameters in real time to address the problem of inability to comprehensively and accurately discern the electrolysis status in the process of rare earth electrolysis and the difficulty to optimize the operation status in time.It is important to study the current efficiency optimization strategy matching with the rare earth intelligent production system to optimize the rare earth production process and improve the rare earth and rare earth product quality.In this thesis,we first establish a current efficiency prediction model based on marine predator optimization neural network（MPA-BPNN）.In order to obtain real-time current efficiency prediction values,the model is trained using actual historical production data;the input variables are determined by mechanistic analysis of the parameters influencing the electrolysis process;and the data quality is improved by data pre-processing methods in order to reduce the interference of data samples to the model.Secondly,a support vector machine（SVM）-based electrolysis tank condition classification model is established,using the electrolysis temperature as the basis for determining the tank condition,excluding the sick tank condition,and dividing three electrolysis temperature intervals corresponding to the corresponding tank condition status of the electrolysis process.At the same time,the model parameters are selected by k-CV method to avoid the overfitting problem.Based on the above two models,the mathematical model for parameter optimization under the process conditions of rare earth electrolysis process is established.And the empirical intervals of controllable parameters are processed to make them constraints,and the multi-objective marine predator algorithm（MOMPA）is used,and the fitness function is calculated to find the optimal solution for the current efficiency and the corresponding parameter sets to obtain the optimal technical condition parameter sets.For the abnormal tank condition and sick tank situation of rare earth electrolysis process,this thesis uses expert experience method and fuzzy set theory to qualitatively analyze and quantitatively judge them,and timely adjust the relevant technical condition parameters to establish the whole set of current efficiency optimization strategy.The experiments show that the optimization strategy proposed in this thesis has effectiveness and timeliness.After using this optimization strategy for guiding adjustment and control,the average current efficiency is improved by 6.79%and the average energy consumption of electrolytic cell is decreased by 164.79 k W·h·t^-1,which can stabilize the current efficiency in the ideal range and effectively avoid abnormal cell conditions and sick cells,while improving the universality of rare earth production system.