锂电池叠片机隔膜纠偏神经近似内模及迭代学习复合控制

A Composite Control of Neural Approximation Inverse and Iterative Learning for Rectifying Separator Film Deviation in Lithium Battery Laminated Machine

为了解决叠片过程中隔膜对齐度较差的问题,采用神经近似内模和迭代学习控制相结合的方法设计控制器来改进隔膜的纠偏效果,提出一种神经网络近似内模及迭代学习复合控制的隔膜纠偏控制算法。首先针对影响隔膜对齐度的复杂特性导致难以用物理数学模型去描述纠偏过程的问题,采用神经网络的优秀的非线性逼近能力建立纠偏系统的神经网络模型。其次为了提升系统的鲁棒性以及避免系统模型的非仿射非线性特性,采用一种神经近似内模对纠偏系统进行控制,仿真表明神经近似内模对纠偏系统能取得较好的控制效果,但是对周期性扰动的抑制能力有限。然而在锂电池叠片过程中,速度和张力的规律性变化会对隔膜偏移误差产生周期性的干扰。最后将迭代学习控制引入到神经近似内模控制中以应对纠偏系统的周期性扰动,仿真表明引入迭代学习控制后,纠偏系统的周期性扰动得到有效地抑制。试验结果表明所提出的纠偏控制算法可以有效地提升锂电池叠片机放卷系统的隔膜对齐度。

Aiming at the problem that the separator film has a poor alignment in the manufacturing process of square lithium batteries using the laminated machine,a novel composite control algorithm is proposed by combining neural network approximate internal model and iterative learning control.Firstly,because the poor alignment is caused by many unknown factors in the manufacturing process of laminated lithium battery,it is impossible to describe the process using a physical mathematical model.A neural network is built to describe the process.Then,in order to enhance the robustness and avoid the non-affine nonlinearity,a neural approximate inverse control is applied to control the process,and the simulation results show that a good performance is obtained in the absence of periodic perturbations.However,the periodic changes of the separator film velocity and tension lead to bring about a periodic disturbance in the process.Finally,the iterative learning control is used to deal with periodic disturbances in the batch manufacturing process of lithium batteries.Experiments demonstrate the composite control algorithm can effectively improve the alignment in the manufacturing process of square lithium using the laminated machine.