Summarize of Lithium Battery State of Health Estimation Method

Summarize of Lithium Battery State of Health Estimation Method
Abstract：Battery management system (BMS) is indispensable for battery technology, and the estimation of battery state of health (SOH) is one of the most important functions of BMS. SOH can provide operation personnel with available capacity and ageing state of the battery, and thus corresponding control strategies can be made or adjusted according these informations. In this paper, definition of SOH is described; and the essential reason for lithium battery aging and capacity fade is expounded; as the main task, the commonly used and studied SOH estimation methods are summarized, general concept of each method is briefly introduced, advantage and drawback for each method is also discussed.
Keywords: electric vehicle; lithium battery; aging causes; energy storage; state of heath estimation
Introduction
In the face of global warming and the decline of fossil energy reserves, the development of low-carbon green energy has attracted worldwide attention. Electric vehicles use electric energy as power source to achieve zero carbon dioxide emissions, so they are favored by countries [1-3]. However, at present, the electric vehicle technology still has the problems of short mileage, high cost and poor safety, which restricts the development of electric vehicle. The root cause of these problems is the immature battery technology [4-5]. Besides battery materials and production process, battery management system BMS is also an important part of battery technology.
BMS is an important link to ensure the normal operation of battery pack. Its main task is to provide accurate and real-time battery status information for operators, to ensure that the battery pack works in a safe range, and to intervene when the working state of the battery is abnormal. Eliminate safety risks in advance and prolong battery life. One of the main functions of the BMS is to accurately estimate the health status SOH of each unit in the battery group. With SOH as a reference, the operator can adjust the charge and discharge control strategy of the battery pack in time. And the aging monomer repair or replacement. Therefore SOH estimation technology is of great significance to enhance BMS performance, improve battery energy storage technology and ensure the safe operation [6] battery power supply system as a whole.
Battery Health Status Definition
A study of the definition SOH battery health began in the late 20th century. The rapid development of batteries has promoted the progress of related technology, SOH estimation and battery aging mechanism are gradually [7] by relevant researchers. The QC/T743-2006 of the automobile industry standard issued by China in 2006 clearly stipulates the conditions for the end of lithium battery life [8]: the available capacity attenuates to 80% of the standard capacity. Can be understood as if the lithium battery SOH reduced to 80%, The battery life ends. Currently, A generally accepted SOH in the industry is defined as
SOH=Q_aged/Q_rate (1)
The Qrate is the rated capacity or nominal capacity of the new battery when it leaves the factory, and the Qaged is the actual available capacity of the battery after it is put into use, and its value will decrease with the increase of charge and discharge cycle times. Although the definition is simple, it is difficult to estimate the accurate Qaged in the normal operation of the battery.
Causes of capacity decay and aging of lithium batteries
Among many types of batteries, lithium battery has become the most promising energy storage [9] because of its long life, high specific energy, good discharge, high safety and small volume. In order to find the main factors of aging of lithium-ion batteries, it is necessary to analyze the microscopic mechanism in the aging process of lithium-ion batteries. Many literatures have studied the capacity attenuation and internal resistance increase of lithium battery in charge and discharge cycle, and understand these phenomena based on the electrochemical mechanism of battery [10-11].
The results show that the aging process of lithium battery mainly occurs at the interface between positive, negative, electrolyte and electrolyte. At present, researchers generally believe that the structural changes at the interface between negative electrode and electrolyte of lithium battery, such as the formation and thickening of solid electrolyte membrane SEI (solid electrolyte interface), are one of the main causes of battery aging. However, from another point of view, in the subsequent charge-discharge cycle, the existence of SEI film can inhibit the further corrosion of electrolyte to negative electrode. J. Vetter and so on, in the process of recycling, the continuous thickening of the SEI film makes the internal resistance rise, which leads to the decrease of the power characteristics of the lithium battery, while the continuous dissolution of lithium metal in the electrode and the gradual loss of active substances lead to the [12] decrease of the energy characteristics of the lithium battery. furthermore, the decomposition of the SEI film at high temperature, the deposition of lithium ions inside the negative electrode during low temperature charging, and the growth of lithium dendrites can lead to the decline of the negative electrode of the battery.
Changes in electrode materials and electrolytes for lithium batteries also affect battery life. According to Wu et al, the main reason for the decrease of reversible capacity of lithium battery is the gradual dissolution of cathode material, phase transition of electrode material and decomposition of electrolyte. The American John NewmNN Research Group and the Michael A. Roscher of the University of Aachen in Germany concluded that the variation of the crystal structure of the positive electrode was obtained by establishing the phase transition model, resulting in an increase in the battery internal resistance and a decrease in capacity [13–14]. For common LiFePO4/C batteries, extremely LiFePO4, this metal oxide has internal stress during lithium ion deintercalation, This may lead to poor crystal stability, The penetration of the cathode/electrolyte interface may change after the reaction between the cathode and electrolyte; Besides, the decomposition and corrosion of inactive substances such as binder incorporated in cathode materials, conductive agents can also lead to deterioration of cathode properties. Overall, for negative batteries, formation and growth of SEI film, corrosion of electrolyte to negative electrode, deposition of lithium metal, decomposition of binder and corrosion of collector are the main causes of electrode aging. For the battery cathode, the lattice structure collapse of the cathode material caused by internal stress, the pulverization and shedding of the active material, the corrosion of the electrolyte to the electrode, the deposition of the new phase of the side reaction, the reaction of the conductive agent, the decomposition of the binder and the corrosion of the collector are the main factors of the electrode aging.
SOH estimation methods
SOH influenced by many factors, and related to battery aging, Many of these parameters are difficult to measure in real time, As a result, relative to the state of charge SOC, SOH estimation is more complex and difficult. At present, there are still many problems in the estimation of lithium battery SOH, Mainly reflected in two aspects :
lack of simple and effective figure 1 lithium-ion battery basic structure Fig.1 Basic structure of lithium battery negative positive electrode electrolyte discharge charging Li electrolyte diaphragm V Zhang Jinlong. The methods of estimating the health status of lithium battery are summarized. The methods of analyzing the aging process of the battery are summarized in the Journal of the 70th phase of the total;
Battery SOH online estimation method is not mature. The real-time performance is poor. and this is also the main research direction [14] related researchers. At present SOH estimation methods can be divided into two categories.
Characteristic method
The SOH is estimated based on SOH evolution of characteristic parameters in the aging process of battery.
Internal resistance analysis
Internal resistance analysis is the most common analysis of SOH based on DC resistance or AC impedance. The internal resistance analysis method takes the battery internal resistance as the main characterization of the battery life. With the aging of the battery and the decrease of the capacity, the battery internal resistance will increase gradually. Therefore, in order to estimate SOH, the corresponding relationship between internal resistance and SOH should be established. Then the SOH. of the battery is located by accurate measurement or estimation of internal resistance. The common method of internal resistance acquisition is pulse method, that is, the current pulse is used to excite the battery, and the internal resistance is estimated according to the change of current and corresponding terminal voltage combined with Ohm’s law and polarization curve fitting. In addition, particle filter, recursive least square RLS or Kalman filter are often used to identify the internal resistance of the battery.
University of Colorado Gregory L. Plett and others measure the internal resistance of the battery and the relationship between the change of electricity and the battery SOH by EKF algorithm, And then estimate the SOH[16]; of the battery Baladizon University of India Neeta Khare used statistical modeling to determine the mapping between battery internal resistance and SOH, Then realize SOH estimation [17]; Dai Haifeng et al analyzed the relationship between internal resistance and current rate, discharge depth and ambient temperature under cyclic charge-discharge conditions, Using the equivalent circuit model to simulate the internal operation of the battery, Then determine the dynamic resistance parameters, And the impedance and the SOH are approximately linear, Through these steps, Finally, SOH[18]; can be obtained by looking up tables And J.Remnilinger and so on based on the battery equivalent circuit model using the linear least square algorithm to obtain the equivalent battery internal resistance parameters, And used for SOH estimation, The method also introduces temperature factors to correct the error of SOH estimation [19]. The difficulty of SOH estimation of internal resistance method lies in the extraction of the relation between SOH and internal resistance mapping, Especially when considering SOC、 temperature and rate factors, The difficulty of extracting feature relationship will increase significantly. In addition, the extracted characteristic relationship is only applicable to a brand of the same type of battery, Poor versatility.
Electrochemical impedance spectroscopy
electrochemical impedance spectroscopy is similar to the internal impedance EIS (electrochemical impedance spectroscopy) method. EIS is the AC impedance spectrum of the battery measured at different frequencies. The impedance spectrum can simulate the electrochemical parameters inside the battery, By monitoring changes in these parameters, Battery life and SOH. can be estimated. The basic idea of this method is to measure the impedance spectrum curve at different aging stages of the battery, And then the EIS curve and the battery equivalent circuit model parameters, Then the SOH is located and according to the relationship between the model parameters (such as series resistance, load transfer resistance and Warburg coefficient, etc[20]. Some studies directly establish the relationship between EIS curve eigenvalues and battery SOH to realize SOH estimation. The A. Zenati of the University of Bordeaux in France applied fuzzy logic to battery SOH estimation, Firstly, they analyzed the aging of batteries at different temperatures. And then the EIS curve is extracted, Drawing Nyquist curves at a wide range of frequencies based on previous experimental data, Finally, the SOH [21]. is obtained by using the curve with fuzzy logic Literature [22] In view of the complex problem of EIS measurement in practical application, an empirical capacity estimation model based on Coulomb’s law is proposed to simulate the capacity loss caused by polarization and concentration difference polarization. At the same time, particle filter is used to track the running state of the system and obtain the parameters of empirical degradation model, and then the residual life and SOH. of the battery are obtained EIS curve contains abundant detail information and can give detailed description of battery impedance. it is suitable for estimating battery aging state and SOH. however, EIS measurement is complex and requires special instruments, which can only be applied offline. In addition, similar to internal resistance method, EIS method is not suitable for battery production, design and process improvement.
Differential analysis
Differential analysis is another popular feature based on differential capacity DC, dQ/dV or differential voltage DV, dQ/dV, SOH estimation of characteristics, such methods are also called incremental capacity IC or incremental voltage IV method [23]. The DC or DV characteristic curve of the battery is closely related to the microelectrochemical mechanism of the battery, The characteristics of peak potential, peak sharpness, asymmetry and peak area of the DC curve are directly related to the thickness of the SEI film, the stress state between the center points of the electrochemical reaction, the reversible degree of the electrode reaction and the content of the active material, According to the battery in different old , And so on: the DC curve set of lithium battery health state estimation method, The aging feature of the battery can be extracted, And then realize the estimation of SOH.
There is also a way to calculate the voltage value points and the corresponding times in the voltage detection data, and draw the probability density curve. Under the same conditions, the more times a voltage appears, the higher the peak value. With battery aging, the corresponding voltage changes, so the mapping between peak voltage and battery SOH can be established.
Other feature analysis
other feature analysis methods also have some other feature-based SOH estimation methods, such as Sun Yuhua of Kaohsiung University of Science and Technology in Taiwan, who directly uses the discharge curve of the battery as the basis for measuring the health state of the battery. With the aging of the battery, the discharge curve will also be distorted and deformed. Based on this phenomenon, the relationship between the characteristic value of the discharge curve and the SOH is established, and the SOH estimation is realized. this method can directly diagnose the aging state of the battery and does not require a complex algorithm; however, in practice, the battery discharge mechanism is random, so this method can not estimate the battery SOH, online can only be an off-line means. The SOH, of the open circuit voltage after the discharge cut-off is mainly based on the recovery rate of the open circuit voltage. Although this method is simple, its SOH estimation accuracy is not high.
Model Method
Aging mechanism model
The most fundamental model for SOH estimation is the aging mechanism model. Similar to the electrochemical model in the SOC estimation, the aging mechanism model also analyzes the physical and electrochemical processes inside the battery, and mainly focuses on the analysis of the aging process of the battery.
At present, the identification methods of battery aging mechanism can be divided into destructive and non-destructive. The destructive means are to completely disassemble lithium batteries with different aging degrees to obtain samples of battery internal materials, and to analyze the relationship between the aging degree parameters and the residual capacity of lithium batteries. The aging mechanism of the battery is extracted by a large amount of data obtained, and the aging model of lithium battery is established. This method mainly includes scanning electron microscope, atomic force microscope, X ray near-edge structure spectrum and Auger electron spectrum. Although this kind of analysis method can directly and accurately give the internal aging situation of the battery and simulate the internal mechanism of most batteries, batteries, but the workload is large, the time consuming, the model is more complex and professional, most of them can only run in the laboratory. It is difficult to realize in practical engineering application. The non-destructive identification method focuses on finding the appropriate aging degree characterization parameters, and establishes the corresponding relationship between these parameters and the aging degree of the battery, and then obtains the aging mechanism model. Similar to the feature method, the SOH estimation accuracy is worse than that of the destructive method, but the workload is small, which is one of the mainstream development directions at present. The technical difficulty is the location and acquisition of characteristic parameters.
The temperature effect of the Arunis model is obvious due to the attenuation of battery capacity, The higher the temperature, The greater the battery capacity, Therefore, more literature focuses on the effect of temperature on battery durability. Arunis studied these chemical reactions in 1880, Based on a lot of data, Proposing an acceleration model.
ζ=Ae^(E⁄KT) (2)
ζ is a life characteristic, such as median life, Average life, etc; A is constant, A>0; E activation energy, Related to materials, eV; K is the Boltzmann constant, 8.617×10-5eV/℃; T is absolute temperature. So the E/K unit is ℃, E/K is also called activation temperature.
It is also common to use electrical stress (such as voltage, current, power, etc.) as acceleration stress in the model of inverse power law. Increasing stress levels can promote early product failure, It’s physically proven by a lot of test data, Some life characteristics of the product are related to stress, The C is a normal number related to activation energy; v stress, Often taken as voltage.
ζ=Av^(-C) (3)
Other models based on fatigue strength models for metal materials, Describe the stress - SOH curve of the battery, By analyzing the diffusion process in the spherical particle model, The lumped parameter (product of SOC increment and SOC mean) is used as the mechanical stress during the cycle. [24]. J. of the University of California Christensen Based on the theory of porous electrode and solid phase diffusion, a mathematical model describing the change of volume, lithium-ion concentration and stress of electrode spherical particles during embedding process of embedding / stripping is constructed. The model can predict the change of electrode crack in,
Probability model
Literature [27-28] combines the battery equivalent circuit model with probability analysis methods (such as Bayesian regression and classification algorithm) to describe the aging and capacity attenuation process of the battery, and verifies the model through experiments. The time interval corresponding to the medium voltage drop sequence during battery capacity and discharge process is [29] used as the health index of the battery. Based on this index, a probabilistic monotone echo state network MONESN (monotone echo state network) algorithm is used to track the nonlinear aging process of the battery. Probability model has two advantages :
only need to carry out partial charge and discharge test of battery to realize SOH estimation.
this method is relatively simple and easy to realize, but the extraction of feature mapping needs to consider many factors, and the workload is large.
Data-Driven Method
This method takes the battery test data as the original sample. The evolution of battery performance during battery aging is discovered by some mechanism, and then this law is used in SOH estimation. This method does not require a systematic analysis of the aging mechanism of the battery, High attention in the industry; But the data-driven SOH estimation algorithm also has some limitations, Because the driving data obtained in the early stage is often uncertain and incomplete. This will inevitably lead to errors. These methods include autoregressive AR model, neural network, support vector machine SVM and Gaussian process regression.
AR model
Approach to battery SOH estimation, AR model uses the combination of linear function and random error of the observation value of the historical state of the system to predict the future performance state. First, the model parameters are determined by nonlinear optimal algorithm, and then build a prediction model, And then the current SOH. of the battery is obtained by extrapolating the system state to the failure threshold .The advantage of using AR model for SOH estimation is that it has less computation and less complexity, The disadvantage is that the accuracy of the results will become worse over a long time. Long B. The attenuation process of the available capacity of the battery is tracked by the AR model, and the particle swarm optimization PSO (particle swarm optimization) algorithm is used to determine the model order, and the online estimation [30] of the aging state of the lithium battery is completed.
NN, SVM and fuzzy methods
The NN, SVM and fuzzy methods are nonlinear, It also has strong generalization ability and parallel processing ability, Suitable for analog batteries, The generality is also strong. A disadvantage of the NN method is that it requires a large amount of comprehensive sample data to train the system, and the estimation error is greatly affected by the training data; In addition, Given its own open-loop structure, NN it’s hard to do online training, while some people combine KF with NN to achieve closed-loop structure NN to improve the adaptive [31] of the algorithm to battery aging, But the complexity of the whole algorithm is greatly increased, It may also lead to overfitting and other problems. SVM like NN. Besides having NN advantages, SVM can also avoid NN structural selection and local minima problems, But SVM also need a lot of sample data for training, And very sensitive to missing data. Fuzzy logic is to describe the knowledge and experience of experts as machine-readable control rules. Thus, the control of the system is realized by computer simulation. Fuzzy control does not depend on the exact model of the system, Robust to the change of model parameters, Like NN and SVM. The establishment of fuzzy rule base requires a lot of experience knowledge and data. And once the control rules are established, You can’t adjust online, Not universal.
Conclusion
Conclusion this paper reviews the aging, capacity attenuation and SOH estimation of lithium batteries, At present, the main problems of SOH estimation technology can be summarized as follows :
the essence of battery aging and capacity attenuation is complex and multi-factor electrochemical and physical processes, And the real-time electrochemical parameters inside the battery are difficult to accurately measure, It is difficult to describe the process accurately and accurately.
The current SOH estimates are mainly for battery testing under specific operating conditions. The actual working conditions may be quite complex; For example, when considering subjective and objective factors, Electric vehicle battery pack conditions will be very random. As a result, SOH estimates have become more difficult.
At present, the generality of SOH estimation technology is poor.
Even for lithium batteries alone, Because battery materials and production processes are different, For different manufacturers and brands, Its performance may differ greatly; Besides, Even with the same brand and the same model, There may also be differences in capacity and voltage characteristics between monomers, And with the charge - discharge cycle, This monomer difference will gradually increase, Therefore, How to design a SOH estimation strategy with strong generality and adaptive ability of monomer difference, Also is the BMS domain urgent need to solve the key question.

References:
Zhang Jinlong, Wei Yanjun, Li Xiangli, et al. Battery SOC estimation based on online parameter identification [J].Transactions of China Electrotechnical Society, 2014, 29(1):23-28.
Ma Zeyu, Jiang Jiuchun, Zhang Weige, et al. Research on path dependence of large format LiMn2O4 battery degradation in thermal aging[J]. Transactions of China Electrotechnical Society, 2014, 29(5):221-227
Aung H, Soon Low K, Ting Goh S. State-of-charge estimation of lithium-ion battery using square root spherical unscented kalman filter in nanosatellite[J]. IEEE Transactionson Power Electronics, 2015, 30(9):4774-83.
Shi Wei, Jiang Jiuchun, Li Suoyu, et al. Research on SOC estimation for LiFePO4 Li-ion batteries[J]. Journal of Electronic Measurement and Instrument, 2010, 24(8):769-774.
Wei Kexin, Chen Qiaoyan. States estimation of Li-ion power batteries based on adaptive unscented kalman filters[J]. Proceedings of the CSEE, 2014, 34(8):445-452.
Jun M, Smith K, Graf P. State-space representation of Liion battery porous electrode impedance model with balanced model reduction[J]. Journal of Power Sources, 2015,273:1226-1236.
Dai Haifeng, Zhou Yanxin, Gu Weijun, et al. Review on life studies of traction Li-ion batteries in electric vehicles[J]. Chinese Journal of Power Sources, 2014, 138 (10):1952-1954.
National Development and Reform Commission. QC/T 743-006: Lithium ion batteries for electric vehicles[S]. Beijing: National Development and Reform Commission, 2006.
Walz K A, Johnson C S, Genthe J, et al. Elevated temperature cycling stability and electrochemical impedance of LiMn2O4 cathodes with nanoporous ZrO2 and TiO2 coatings[J]. Journal of Power Sources, 2010, 195(15):4943-4951.
Karsten M，Daniel T，Lars G，et al. Optimizing BMS operating strategy based on precise SOH determination of lithium-ion battery cells[J]. Lecture Notes in Electrical Engineering, 2013, 192(4):807-819.
Kim J, Lee S, Cho B, et al. Discharging/charging voltage temperature pattern recognition for improved SOC/capacity estimation and SOH prediction at various temperatures[J]. Journal of Power Electronics, 2012, 12（1）:1-9.
Vetter J, Novak P, Wagner R, et al. Aging mechanisims in lithium-ion batteries[J]. Journal of Power Sources, 2005, 147(1/2):269-281.
Roscher M A,Vetter J,Sauer D U. Characterization of charge and discharge behavior of lithium ion batteries with olivine based cathode active material[J]. Journal of Power Sources, 2009, 191(2):582-590.
Srinivasan V, Newman J.Discharge model for the lithium iron-phosphate electrode[J]. Journal of the Electrochemical Society, 2004, 151(10):A1517-A1529．
Xu Xinmin, Wang Lian, Shi Huiling. Research on battery aging mechanisms based on electrochemical impedance spectroscopy[J]. Chinese Journal of Power Sources, 2015,139(12):2579-2583.
Gregory P. State-of-charge estimation based on microcontroller-implemented sigma-point Kalman filter in a modular cell balancing system for Lithium-Ion battery packs [C]. IEEE Conference Publications, 2015:1-7.
Khare N, Singh P, Vassiliou J K. A novel magnetic field probing technique for determining state of health of sealed lead-acid batteries[J]. Journal of Power Sources, 2012, 218（218）:462-473.
Dai Haifeng, Wei Xuezhe, Sun Zechang. A new SOH prediction concept for the power lithium-ion battery used on HEVs.[C]. IEEE Vehicle Power and Propulsion Conference VPPC, Dearborn, 2009.
Remmlinger J, Buchholz M, Meiler M, et al. State-of health monitoring of lithium-ion batteries in electric vehicles by on-board internal resistance estimation[J]. Journal of Power Sources, 2011, 196 (12): 5357-5363.
Reinhold K, Robert K, Zilberman I, et al. Electrochemical impedance spectroscopy for online battery monitoring-power electronics control[C]. 16th European Conference on Power Electronics and Applications, Lappeenranta, 2014.
Zenati A, Desprez Ph, Razik H. Estimation of the SOC and the SOH of li-ion batteries, by combining impedance measurements with the fuzzy logic inference［C］. IEEE Industrial Electronics Society, 2010, 36:7-10.
Saha B, Goebel K. Modeling Li-ion battery capacity depletion in a particle filtering framework[C]. Annual Conference of the Prognostics and Health Management Society, 2009.
Dubarry M, Truchot C, Liaw B Y. Synthesize battery degradation modes via a diagnostic and prognostic model[J]. Journal of Power Sources, 2012, 219(12):204-216.
Narayanrao R, Joglekar M M, Inguva S. A phenomenological degradation model for cyclic aging of lithium ion cell materials [J]. Journal of the Electrochemical Society, 2013,160(1):125-137.
Christensen J, Newman J. Stress generation and fracture in lithium insertion materials[J]. Journal of Solid State elecrochemistry, 2006, 10（5）:293-319.
Li J, Barillas J K, Guenther C, et al. Sequential Monte Carlo filter for state estimation of LiFePO4 batteries based on an online updated model[J]. Journal of Power Sources,2014, 247(3):156-162.
Saha B, Goebel K, Poll S, et al. An integrated approach to battery health monitoring using Bayesian regression and state estimation[C]. IEEE Autotestcon, Baltimore, MD, 2007.
An D, Choi J H, Kim N H. Prognostics 101: A tutorial for particle filter-based prognostics algorithm using Matlab[J]. Reliability Engineering & System Safety, 2013, 115(1):161-169.
Liu Datong, Xie Wei, Liao Haitao, et al. An integrated probabilistic approach to lithium-ion battery remaining useful life estimation[J]. IEEE Transactions on Instrumentation and Measurement, 2015, 64（3）:660-670.
Long Bing， Xian Weiming， Jiang Lin， et al. An improved autoregressive model by particle swarm optimization for prognostics of lithium-ion batteries[J]. Microelectronics Reliability, 2013, 53（6）:821-831.
Shahriari M, Farrokhi M. Online state-of-health estimation of VRLA batteries using state of charge[J]. IEEE Transactions on Industrial Electronics, 2013, 60(1):191-202.