Research On Reconstruction Method Of Mechanism Model And Key State Estimation For Vehicle Power Battery

The development of electric vehicles(Evs)is an important way to implement the electrification of transportation in China,which can effectively promote the achievement of energy conservation and emission reduction and double-carbon goals.Power batteries are widely used in power storage systems of EVs because of their low cost and high energy density.In recent years,with the increasing requirements for the safety of EVs,the online detection and evaluation of the operating status and reliability of power batteries has become one of the important challenges in development and application of the energy storage management system of EVs.However,the power battery is a con,plex nonlinear system,and reasonable modeling and accurate state estimation still face lots of challenges.According to the actual application requirements,this paper takes the power battery as the research object,combining the relevant experimental data and the Multiphysics simulation platform,and carries out the estimation of battery state of charge(SOC)and heat generation rate(HGR)of EVs based on the mechanism reconstruction model,mainly including the following research work:(1)The construction of Power battery test platform and Multiphysics simulation system.The power battery test platform can analyze the battery characteristics through routine testing,and collect experimental data under different conditions for the verification of model development and algorithms.In addition,the Multiphysics simulation system can establish the connection between the battery electrochemical reaction and the external characteristics,and provides a basis for the accurate analysis of the internal state estimation of the power battery.(2)Aiming at the problem that the existing power battery mechanism model is complex and its real-time performance is poor,this paper proposes a reconstruction method of power battery electrochemical mechanism model combined with the mathematical reduction method to promote its real-time application in the advanced battery management system.For the solid-phase diffusion equations,a hybrid method consisting of Pade approximations and polynomial approximations is used to reduce computational complexity while ensuring the observability of the battery model system.The volumetric averaging method is used to simplify the time-domain distribution of electrolyte concentrations to compensate for the shortcomings of traditional single-particle model.At the same time,combined with experimental and simulation data,the battery mechanism reconstruction model is verified and the observability is evaluated under different operating conditions.The results show that the proposed mechanism reconstruction model has high model fidelity and lower computational complexity,and improves the observability of the battery system,which is conducive to the estimator design based on the battery mechanism model.(3)In view of the lack of physical characteristics and difficult engineering application of existing SOC estimation,this paper combines the battery mechanism reconstruction model and the proportional integral observer from a physical point of view to estimate the SOC of the power battery.Considering the bias of the current sensor in practical applications,a method of SOC estimation is first proposed.And the current error caused by the sensor bias can be eliminated by the proportional integral observer.At the same time,considering that the battery potential of negative electrode has important role in the optimization of the charging strategy,this paper also uses the proportional integral differential observer to make a collaborative estimation based on physical SOC and the negative electrode potential,and realizes the simple and effective estimation method of battery internal states,which is suitable for engineering applications.The results show that the absolute error of estimated SOC is basically maintained within 2%under HPPC test conditions,and the root mean square error of the negative electrode potential can be maintained at 4.31 mV under the US06 conditions,which provides a strong guarantee for the accurate estimation of SOC of power battery from a physical point of view.(4)In view of the problems that the thermal characteristics of power batteries in the discharge process and its influencing factors are difficult to comprehensively evaluate,this paper establishes a mechanism-thermal coupling model under multiple temperature ranges,and studies the thermal characteristics and influencing factors of cylindrical lithium-ion batteries.Firstly,based on the experimental data at ambient temperatures of 25℃ and 35℃,the accuracies of the developed mechanism-thermal coupling model are verified,and the thermogenic mechanism in the battery discharge process is revealed.Secondly,the effects of different discharge rates and positive/negative electrode(N/P)capacity ratio on the thermal characteristics of battery are comprehensively analyzed.The results show that the heat generation production of negative electrode accounts for the main part of the total heat generation production of the battery,and the appropriate N/P capacity ratio can improve the total heat generation of the lithium-ion battery,which is conducive to optimizing the battery thermal management design and follow-up HGR estimation research.(5)Aiming at the problems such as the difficulty of online estimation of the battery HGR,this paper combines the mechanism reconstruction model and the bidirectional long short-term memory(BiLSTM)network to propose a physical information neural network(PINN)-based battery HGR state estimation method.The surface concentrations of the positive and negative electrodes in the mechanism reconstruction model are integrated into the BiLSTM network model as physical features.Meanwhile,defining key hyperparameters with Bayesian optimization algorithms improves the training speed and accuracy of BiLSTM.Based on the battery HGR data from the Multiphysics simulation system,the estimation accuracy of the proposed method is evaluated.The results show that the root mean square error of the PINN-based battery HGR estimation results is 1.146 kW/m~3 under the dynamic stress test condition,which is helpful for the thermal fault detection and early warning of the battery thermal management system.