铝电解过程神经网络预测控制技术应用研究

Application Research on Neural Network Predictive Control Technology in Aluminum Electolysis Process

铝电解是非线性、时变、大时滞过程,受强电场、强磁场、强热场交互干扰,其控制难度大,耗能高。因此,节约电能,提高电流效率,提高铝的产量和质量是铝电解控制系统研究的热点问题。文中在分析铝电解过程工作特性及存在问题的基础上,提出了小波神经网络预测控制方法,该方法将神经网络控制技术与预测技术有机结合,通过对反映氧化铝浓度的槽电阻参数跟踪,实时调整控制器的控制策略,控制氧化铝下料装置下料量,使氧化铝浓度控制在理想值范围,并对系统的硬件和软件进行了设计。实验结果表明:该方法的有效性,具有良好的控制控制性能和节电效果,对提高铝的产量和质量具有重要意义。

Aluminum electrolysis is a nonlinear,time-varying and a large time delay process,which interfered by the interaction of strong electric field,strong magnetic field and strong heat field.So,it is a high energy consumption process and the process control is very difficult.Therefore,the hot issue for the control system is how to save energy,improve the current efficiency,increase the yield and the quality of aluminum electrolysis.A wavelet neural network predictive control method was proposed which based on the analysis of characteristics and problems for the aluminum electrolysis process.The method combined the neural network control technology and forecasting techniques.By tracking the parameter of the cell resistance which reflects the alumina concentration,the controller regulates the control strategy on real-time to make the alumina concentration in an ideal range through controlling the alumina feeding quantity of the feeding device,and the system＇s hardware and software were designed.The experiment results show that the method has a good effective control performance and an energy-saving effect,and has an important significance of increasing the yield and quality of aluminum.