Study On SOH Estimation Of Retired Li-ion Batteries Based On EIS

Reducing carbon emissions to combat adverse climate change has become a global consensus,and the development of green and sustainable new energy is an important way to reduce carbon emissions.With the rapid development of new power systems and new energy vehicles mainly based on new energy sources,energy storage systems based on lithium-ion batteries have been widely used.However,due to the limited number of charge/discharge cycles of lithium-ion batteries due to their under-reactive electrochemical characteristics,a large number of lithium-ion batteries are currently facing retirement and need to be disposed of properly.To maximize and rationalize the use of batteries,they can be reused in scenarios with lower performance requirements and relatively mild operating conditions,reducing the environmental pollution caused by the large number of disposed batteries.However,the state of health(SOH)of retired batteries needs to be determined and assessed to screen them for reuse.To address the problems that most of the battery health states cannot be detected directly but can only be estimated indirectly,and that current estimation algorithms or strategies have limitations in the estimation model,or the physical meaning of the model is unclear,or the model has poor generalization capability and requires large training data to support its accuracy,this paper investigates the electrochemical impedance spectroscopy(EIS)of batteries and its impedance characteristics,and proposes a combination of battery-based model and data-driven approach to achieve SOH estimation of Li-ion batteries:In addressing the problems that the battery health status cannot be directly detected but only estimated by indirect methods,and that the current estimation algorithms have strong limitations of the estimated model,unclear physical meaning of the model,poor generalization ability of the model and large amount of training data,this paper investigates the Electrochemical Impedance Spectroscopy of Li-ion batteries and their impedance characteristics,based on which a combination of cell-based models and data-driven methods for SOH estimation of Li-ion batteries are investigated,with the following research work.(1)The operating principle of lithium-ion batteries is analyzed,and the second-order RC equivalent circuit model is chosen to simulate the non-linear operating characteristics of the batteries;to specifically establish the equivalent circuit model of lithium-ion batteries,the EIS method is introduced to build a simulation experiment of the battery model based on Matlab/Simulink;the correlation between the SOH of the battery and the impedance of the battery characterizing the electrochemical reaction is analyzed,and the influence of temperature and SOC(State of charge)on the impedance characteristics is studied based on Matlab Simulation experiments based on Matlab/Simulink investigated the effects of temperature and SOC on impedance characteristics,which laid the foundation for the identification of equivalent circuit model parameters and SOH estimation of the battery in multiple states.(2)In order to identify the parameters of the second-order equivalent circuit model of the Li-ion battery,the nonlinear least squares method(Nonlinear LSM)for offline identification and the particle swarm optimization algorithms(PSO)for online identification were used to identify the impedance of the Li-ion battery at different frequencies.In addressing the problem that offline identification algorithms require manual processing of data and are cumbersome to apply and that online identification algorithms have large fitting error results,this paper investigates the characteristics of both algorithms and proposes a joint nonlinear least square online PSO algorithm,which is only processed once offline and has high identification accuracy,and the identification results demonstrate its effectiveness.(3)In order to estimate the SOH of the battery,this paper investigates the use of linear regression(LR),which is easier to abstractly map the relationship between variables,and Gaussian process regression(GPR),which is based on a stochastic process and probabilistic modelling of the non-linear relationship,to map the SOH with respect to T,SOC,and RC.The functional relationship model is analyzed and compared between LR and GPR training models for SOH estimation through multiple iterations of the experiment,and the experimental results show that GPR is more effective in estimating SOH for Li-ion batteries.(4)EIS test is the key link to realize parameter identification and SOH estimation.In this paper,we developed our own hardware experimental platform to realize EIS test,obtained EIS data of Li-ion battery under multiple states,then built a hardware-in-the-loop experimental platform,and combined with EIS test hardware to verify the feasibility and effectiveness of the algorithm in this paper,and realized SOH estimation,which has certain engineering application value.The development of this study may provide new ideas and methods for the validity and accuracy of SOH estimation prior to lithium-ion battery gradient utilization.