Research On Fluorinated Aluminum Additives For Aluminum Reduction Cell Based On The Neural Network

To improve the current efficiency, energy consumption has been aluminum electrolytic researchers tireless pursuit, especially in the current situation of the energy shortage, aluminium electrolysis production costs rise. Existing technology conditions isnot able to realize to on-line measure current efficiency,so we usually improve current efficiency by adjusting the process parameters. Daily adding weights of AIF3 is one of the most important process parameters for aluminum electrolysis production. According to the situation of the cell, to reasonablely add AIF3 may in a certain extent adjust molecular ratio and temperature, keep them in an ideal range and keep electrolytic bath stable and highly effective.This is one of the importmant methods to improve the current efficiency and save energy consumption.Due to the aluminum reduction cell is a nonlinear, large delay, high temperature and strong corrosive complex system, it is difficult to measured many process parameters and look for accurate maths model to describe the relations among parameters. These cause the difficult decision of daily adding weights of AIF3. After studying fully aluminum electrolysis process and the neural network theory, the research proposes to find the coupling relationship between cell internal balance and daily fluorinated aluminum additives using BP neural network nonlinear mapping function, then reoccupy this relationship to decide daily fluorinated aluminum additives. To overcome the inherent defect of the neural network-stall at the minimum and slow convergent speed, using genetic algorithm to optimize neural network weights and threshold. Given the process of creation of BP neural network and realized the creation in MATLAB software. Through the test of 20 sets of data collecting from production site network output value tallied well with actual ideal values-the absolute errors within3 and the relative error less than 0.1 According to the actual situation of the aluminum electrolysis production, the effect of error is very small. So this method is verified to achieve good effect.Lastly, developed fluorinated aluminum adding decision-making system software in MATLAB7.0.4,SQL Server 2000 and Visual C++6.0 to realize the calculation of fluorinated aluminum additives online, remote view of data and remote management of data.