Design And Optimization Of Electrode Coating Of Lithium-ion Battery

Lithium-ion battery has wide applications for its high energy density,high power density and long cycle life.Strategies of increasing the thickness and C-rates can separately improve energy density and power density of lithium-ion battery theoretically,while the increase of diffusion resistance will result in diffusion control of the discharge and charge process.In this work,in order to simultaneously improve energy density and power density of lithium ion battery,we introduced macro-pore channel as fast transport paths of lithium ion.In order to efficiently optimize geometrical parameters of macro-pore channel structured electrode,we introduced effective concentration resistance as a comprehensive descriptor to describe its performance,which was based on the assumption of homogeneous reaction rate distribution on electrode when the whole electrode process are governed by electrochemical reaction on the interface between active material and electrolyte.The calculation of effective concentration resistance was based on three different methods:parallel resistor method,semi-analytical solution and analytical solution.The validation of effective concentration can be descriptor of performance of battery was proved through simulation and the geometry parameters of macro-pore channel were optimized based on the effective concentration resistance which calculated by parallel resistor method,we found that the power density can increase by 60%compared to the thin-film battery which possess equal active material load.Besides,hierarchical macro-pore channels were introduced into thin-film lithium-ion battery and the optimization of its geometrical parameters was based on the analytical solution of macro-pore channel structured electrode,we found that the performance of hierarchical macro-pore channel structured electrode was not sensitive to its geometrical parameters in some region.To better predict the value of effective diffusivity of lithium ion in porous space of electrode,we proposed convolutional neural network as an effective method to train and predict its value.There are two steps when predict effective diffusivity of microstructure of porous material through convolutional neural network:the first step is generating 27000 samples of graphs of porous material microstructure and the calculations of ground truth of effective diffusivity of samples were based on finite element method tirough Matlab and Comsol.The second step includes all samples were separated into three parts:24000 samples as training samples,1500 samples as validation samples and 1500 samples as test samples.The training of convolutional neural network was on single NVIDIA K80 GPU of Google Co-laboratory and the overfitting during training was alleviated through Dropout.The test accuracy of trained convolutional neural network can reach 96.7%,which forcefully prove that convolutional neural network can precisely predict the effective diffusivity of microstructure of porous material.