Research On Modeling And Estimation Framework Of State Parameter For Lithium-Ion Battery

As the source of energy for new energy vehicles,lithium-ion batteries are increasingly applied with their high energy density,low self-discharge rate and zero pollution.To ensure the reliability and safety of lithium-ion battery operation,it is crucial to equip an intelligent and efficient battery management system for real-time fault prognostics and health management.Among them,State of Health(SOH)and State of Charge(SOC)for lithium-ion batteries are the core elements of Battery Management System(BMS).However,the existing modeling and mixed estimation implementation of SOH and SOC for lithium batteries still suffer from a series of difficulties in health feature extraction,unsuitable models,low accuracy,and fewer research on mixed estimation.In response to the above problems,this paper presents the research on modeling and estimation framework of state parameter for lithium-ion batteries,and the main research contents and key innovations are summarized as follows:(1)Data mining and correlation analysis are performed between the trend of battery measurable parameters and capacity decay,and then the method that extract battery health characteristics during the charging and discharging stages is constructed.Firstly,four categories of 6-dimensional HFs were extracted from peak height,voltage position,maximum slope and peak integral in the charging phase,and 3-dimensional HFs were extracted from voltage time,temperature time in the discharging phase.Next,the above-mentioned HFs which were extracted from the different stages are optimized by combining the traversal search method and principal component analysis.Lastly,the HFs which can not only to adapt to complex working conditions but do not require parameter identification work are obtained.(2)The SOH estimation framework based on Improved Guassian Process Regression(IGPR)model and mixed models is proposed to describe the nonlinear,dynamic changes and the tendency of capacity regeneration during battery capacity degradation.Firstly,the HFs in the discharge phase are used as input and the reference SOH is used as output to train the IGPR model.Meanwhile,the SOH estimation framework based on a double-exponential empirical degradation model and an IGPR mixed model is constructed.Which can avoid the shortcoming and the divergence of prediction results when single data-driven models deal with small sample prediction problems.Lastly,the two SOH estimation models are validated on two datasets,demonstrating the superiority of the mixed model when facing 50% priori information.(3)To realize the state prediction of lithium-ion battery systems at different time scales,a joint SOC and SOH estimation framework based on first-order equivalent circuit model-improved particle filter-IGPR is proposed.Firstly,the IGPR aging model with HFs in the charging stage as input of Li-ion batteries is constructed.Meanwhile,according to the results of resistive-capacitance parameter identification and SIMULINK validation,the state space model of SOC eatimation is constructed and then combined with an improved particle filtering algorithm to update the SOC of the latter cycle to achieve joint long-term estimation of SOC and SOH.Finally,the accuracy and adaptability of the framework is demonstrated by conducting simulation validation experiments on the Oxford University dataset.