10 1016@j Rser 2020 110015

Renewable and Sustainable Energy Reviews 131 (2020) 110015
Contents lists available at ScienceDirect
Renewable and Sustainable Energy Reviews

journal homepage: http://www.elsevier.com/locate/rser
A comprehensive review of battery modeling and state estimation

approaches for advanced battery management systems
Yujie Wang, Jiaqiang Tian, Zhendong Sun, Li Wang, Ruilong Xu, Mince Li, Zonghai Chen *
Department of Automation, University of Science and Technology of China, Hefei, Anhui, 230027, PR China
A R T I C L E I N F O A B S T R A C T
Keywords: With the rapid development of new energy electric vehicles and smart grids, the demand for batteries is
Battery modeling increasing. The battery management system (BMS) plays a crucial role in the battery-powered energy storage
State estimation system. This paper presents a systematic review of the most commonly used battery modeling and state esti
Energy storage
mation approaches for BMSs. The models include the physics-based electrochemical models, the integral and
Battery management
Lithium-ion battery
fractional order equivalent circuit models, and data-driven models. The state estimation approaches are analyzed
from the perspectives of remaining capacity and energy estimation, power capability prediction, lifespan and
health prognoses, and other crucial indexes in BMS. This present paper, through the analysis of literature, in
cludes almost all states in the BMS. The estimation approaches of state-of-charge (SOC), state-of-energy (SOE),
state-of-power (SOP), state-of-function (SOF), state-of-health (SOH), remaining useful life (RUL), remaining
discharge time (RDT), state-of-balance (SOB), and state-of-temperature (SOT) are reviewed and discussed in a
systematical way. Moreover, the challenges and outlooks of the research on future battery management are
disclosed, in the hope of providing some inspirations to the development and design of the next-generation BMSs.
1. Introduction obtained by direct measurement approach and is easily affected by

environmental temperature, noise, etc. The upsizing of power batteries
Energy storage technology is one of the most critical technology to reduces the representativeness of measured values, and reduces the
the development of new energy electric vehicles and smart grids [1]. predictability of battery states, which makes it hard to accurately esti
Benefit from the rapid expansion of new energy electric vehicle, the mate the internal states of the battery; (3) the inconsistencies of the
lithium-ion battery is the fastest developing one among all existed battery cells directly influence the efficiency of the pack, which in
chemical and physical energy storage solutions [2]. In recent years, the creases the hidden danger of the battery. Some effective safety measures
frequent fire accidents of electric vehicles have pushed electric vehicles on small battery systems have little effect on electric vehicles, and the
to the subject of public opinion, and also put forward high requirements efficient and accurate control of the battery pack is difficult. Therefore
and challenges for battery management technology [3]. As one of the advanced BMSs should be designed to solve the above problems [4,5].
key components of electric vehicles, the lithium-ion battery manage The functional structure diagram of an advanced BMS is shown in
ment system (BMS) is crucial to the industrialization and marketization Fig. 1. The key features of the battery management system is shown in
of electric vehicles. Therefore, developing advanced and intelligent Fig. 2. The basic functions of a BMS include battery data acquisition,
BMSs for the lithium-ion battery packs has become a hot research topic. modeling and state estimations, charge and discharge control, fault
The main technical difficulties restricting the development of battery diagnosis and alarm, thermal management, balance control, and
management technology can be concluded in the following three as communication. Battery modeling and state estimation are key functions
pects: (1) the lithium battery system is highly nonlinear, with multi- of the advanced BMS. Accurate modeling and state estimation can
spatial scale (such as nanometer active materials, millimeter cell, and ensure reliable operation, optimize the battery system and provide a
meter battery pack, etc.) and multi-time scale aging, making it difficult basis for safety management [6].
to accurately modeling; (2) the internal states of the battery cannot be In this paper, the hotspots of modeling and state estimation in battery
* Corresponding author.
E-mail addresses: wangyujie@ustc.edu.cn (Y. Wang), tjq1992@mail.ustc.edu.cn (J. Tian), szd1996@mail.ustc.edu.cn (Z. Sun), change94@mail.ustc.edu.cn
(L. Wang), rlxu@mail.ustc.edu.cn (R. Xu), limince@mail.ustc.edu.cn (M. Li), chenzh@ustc.edu.cn (Z. Chen).
https://doi.org/10.1016/j.rser.2020.110015
Received 11 March 2020; Received in revised form 7 June 2020; Accepted 17 June 2020
Available online 16 July 2020
1364-0321/© 2020 Elsevier Ltd. All rights reserved.
Y. Wang et al. Renewable and Sustainable Energy Reviews 131 (2020) 110015
management are selected to discuss. The basic theory and application voltage, and proposes a partial differential equation for liquid electro
methods of battery system modeling and state estimation are reviewed lyte material conservation and charge conservation, which improves the
systematically. The most commonly used battery models including the simulation accuracy. They also solved the problem of multi-parameter
physics-based electrochemical models, the integral and fractional-order estimation in SPMe and conducted sensitivity analysis on these
equivalent circuit models, and the data-driven models are compared and parameters.
discussed. The battery states including the state-of-charge (SOC), state- Doyle and Newman et al. [14] presented the
of-energy (SOE), state-of-power (SOP), state-of-function (SOF), state-of- pseudo-two-dimensional (P2D) model. The P2D model treats the anode
health (SOH), remaining useful life (RUL), remaining discharge time and cathode of the cell as porous electrodes composed of numerous
(RDT), state-of-balance (SOB), and state-of-temperature (SOT) and their spherical particles, and the spaces between particles are filled with the
estimation approaches are reviewed and discussed. Moreover, the electrolyte. The concentration distribution and potential distribution of
challenges and outlooks of the research on battery modeling and state the lithium-ion in the electrolyte liquid phase and the solid phase of the
estimation are proposed. electrode material particles are described by several coupled partial
The remainder of the paper is organized as follows: Section 2 reviews differential equations (PDEs), so as to comprehensively examine the
the battery modeling techniques. Section 3 provides the state of the art main reaction and side reaction rates inside the cell. Since the P2D
of battery state estimation methods. The challenges and outlooks of the model contains several coupled PDEs, it is necessary to simplify the P2D
research on battery modeling and state estimation are presented in model from the perspective of engineering practice [15,16]. Further
Section 4, and finally, Section 5 concludes the paper. more, some coupled battery models have been proposed to represent
different features of battery. A P2D electrochemical-thermal-capacity
2. Battery modeling fade coupled model was developed in Ref. [17]. Ref. [18] Proposed a
fully coupled electro-chemo-mechanical model for thin-film battery,
The battery models presented in literature mainly fall into the which accounted for several mechanical and electrochemical factors.
following three main categories: the physics-based electrochemical One of the main reasons that the physics-based electrochemical models
models [7], the electrical equivalent circuit models (include the are difficult to apply in real-time applications is due to a large number of
integral-order and fractional-order models) [8,9], and the data-driven unknown variables need to be identified using global optimization
models establish by artificial intelligence algorithms such as the neu methods. Inevitably, they are likely to run into over-fitting or local
ral network [10] and support vector machine [11]. optimization problems. In the absence of accurate and detailed model
parameters, the simulation results of the actual physics-based electro
chemical models are often not ideal.
2.1. Physics-based electrochemical models
The single-particle (SP) model is the most mature simplified model 2.2. Electrical equivalent circuit models
for the physics-based electrochemical models [12]. In the SP model, a
single particle is used to represent the concentration distribution of The electrical equivalent circuit models have gained much interest in
lithium-ion in the electrode. The SP model can be used to investigate the real-time applications due to their simplified model structure and easy to
primary performance and the effect of solid-phase diffusion of the be identified. The electrical equivalent circuit models use electrical
electrodes. However, the accuracy of the SP model is relatively low. components to mimic the battery behavior. In literature, the equivalent
Grandjean et al. [13] established the SPMe model based on the SP circuit models can be divided into two groups: the integral-order models
model. The model considers the influence of electrolyte on the output and fractional-order models.
Fig. 1. Functional structure diagram of an advanced battery management system.
2
2.2.1. Integral-order equivalent circuit models

The simplest integral-order equivalent circuit model is the Rint
model which is composed of an ideal voltage source in series with a
resistor [19]. The model structure of the Rint model is simple, but the
polarization and diffusion phenomena are not considered. Liaw et al.
[20] presented the first-order resistor-capacitor (RC) model which is
capable of simulating charge and discharge behavior of lithium-ion
batteries by employing one RC network. This model can simulate the
battery behavior with high fidelity as validated by experimental results.
Xia et al. [21] utilized the second-order RC equivalent circuit model to
approximate the battery dynamic performance and the voltage error is
less than 40 mV. Andre and Sauer et al. [22,23] presented the ideal
impedance spectrum of a lithium-ion cell and an equivalent circuit
model with three RC networks was proposed. There is also literature
considering the hysteresis behavior of the open-circuit voltage (OCV) to
enhance the model accuracy [24].
For the integral-order equivalent circuit models, the input/output
relationship of the cell is easy to derive and the models involve minor
parameters. Therefore, the most widely used methods for on-line
parameter identification are the recursive least-squares method. Wang
et al. [25] proposed a co-estimator to estimate the model parameters and
battery state. In Ref. [26], Zhang et al. proposed a novel parameter
identification method considering the electrochemical properties. The Fig. 3. Fractional-order equivalent circuit model and EIS diagram of a lithium-
presented method combines the electrochemical properties and equiv ion battery.
alent circuit to study lithium-ion battery performance in different con
ditions, which better reflects the battery interior mechanism. Brand characteristic on the whole frequency range. To overcome this problem,
et al. [27] proposed a multi-objective genetic algorithm for parameter the constant phase elements (CPEs) can be used instead of the capacitors
extraction of the equivalent circuit model. The results indicate that the in RC networks. Therefore the fractional-order equivalent circuit models
method can accurately estimate the voltage of different types of are proposed [30].
batteries. In Ref. [31], Yang et al. presented a simplified fractional order
impedance model. The genetic algorithm is used to identify the model
2.2.2. Fractional-order equivalent circuit models parameters and the voltage error is within 0.5%. Zou et al. [32] pre
The electrochemical impedance spectroscopy (EIS) [28] and the sented the non-integer order model derivatives and the particle swarm
bode plot [29] are useful tools for analyzing the battery characteristics optimization algorithm was employed to identify the model parameters.
and establishing the model structure. Fig. 3 shows the EIS diagram for The proposed fractional-order equivalent circuit model showed good
the impedance of a lithium-ion battery. In the middle-frequency, the accuracy and robustness. In recent years, the fractional-order models are
battery’s Nyquist curve is not a standard semicircle, which means that widely used for battery state estimation and reliability analysis. Mu et al.
the standard RC network is not appropriate to simulate the battery [33] and Xiong et al. [34] employed the fractional-order equivalent
Fig. 2. Key features of the battery management system.
3
circuit model and Kalman filter to estimate the SOC of the lithium-ion 3.1. Remaining capacity and energy estimation
batteries. Wang et al. [35] proposed the fractional-order models for a
battery and supercapacitor hybrid system. Hu et al. [36] and Tian et al. Physically, discharging a cell moves lithium from the negative
[37] employed the fractional-order models to estimate the battery SOH electrode to the positive electrode, and the charging process does the
and degradation state. In Ref. [38], the fractional-order model was used opposite. Electrochemically, the cell state-of-charge (SOC) is positively
for the battery external short circuit fault diagnosis. A comparison of the related to the average concentration of lithium in the negative-electrode
benefits and drawbacks of the integral-order and fractional-order solid particles, and negatively related to the average concentration of
models is shown in Fig. 4. lithium in the positive-electrode solid particles [44]. It should be noted
that the cell voltage depends on electrode particle surface concentra
2.3. Data-driven models tions but SOC depends on average particle concentrations, which makes
determining SOC using only voltage problematic. This is why the battery
The data-driven methods have been widely used in the development capacity is difficult to obtain and measure directly. Another question for
of battery models. Wang et al. [39] presented a thermal-electrical the BMS to answer is how much energy is available in the battery pack.
coupled model based on neural network. The effectiveness of the pro This question is most important for the electric vehicle because it is a
posed approach was validated by different battery tests. Zhang et al. fundamental input to vehicle range calculations which can tell the
[40] proposed a data-driven state estimation approach by combining the drivers how far they can drive. Therefore, the remaining energy or
neural network with particle filtering. Dong et al. [41] proposed a state-of-energy (SOE) estimation is crucial.
data-driven battery model based on wavelet-neural-network. In
Ref. [42], the Stacked Denoising Autoencoders algorithm and the 3.1.1. Review of state-of-charge estimation approaches
Extreme Learning Machine algorithm were combined to form a big
data-driven lithium-ion battery model, which considered the impact of 3.1.1.1. Look-up table method. The look-up table method is convenient
temperature. Although the data-driven approaches have good perfor and straightforward, which measures SOC based on the mapping rela
mance to nonlinear problems, they are easily influenced by training tionship between characteristic parameters (such as impedance spec
datasets and training methods. In addition, the dynamic simulation troscopy, internal resistance, OCV, etc.) and SOC [45].
technology is also used in battery modeling. Vigneshwaran et al. [43]
presented a three-dimensional kinetic Monte Carlo model to reveal the ● Open-circuit voltage (OCV) look-up table method. The OCV is the
law of structural evolution of the dissolution/precipitation reaction of battery voltage after a long time of resting without load, which has a
solid sulfur and lithium sulfide during the discharge of lithium-sulfur nonlinear relationship between SOC [46]. The detailed imple
batteries. The kinetic Monte Carlo (kMC) model can approximate the mentation process of the method is shown in Ref. [47]. The OCV is
dynamics of the battery at a longer time scale, so it can effectively measured under different SOC conditions and made into the
predict the structural evolution of Li–O2, slurry redox flow batteries and OCV-SOC table. In practice, the SOC is determined by measuring
other types of batteries. OCV and OCV-SOC table. The method is simple and easy, but the
battery needs to be standing for a long time to ensure that the
3. Battery state estimation measured voltage is equal to OCV. Considering the influence of
temperature, material, and aging on the relationship between OCV
The battery is a complex nonlinear system with multiple state vari and SOC, these factors are added to the OCV-SOC table [48]. Pecht
ables, therefore the accurate estimation of battery states is the key to et al. [49] established an off-line OCV-SOC-Temperature table to
battery management and the basis of battery control. This section sys infer battery SOC. Besides, the hysteresis effect of OCV is also an
tematically summarizes the theoretical methods of battery state esti essential factor affecting the accuracy of OCV [50]. The hysteresis
mation from the following four aspects: remaining capacity & energy effect can be identified as the difference of OCV under the same SOC
estimation, power capability prediction, lifespan & health prognoses, during charging and discharging, especially for LiFePO4 batteries
and other important indexes in BMS. [51]. Therefore, an inaccurate OCV-SOC relationship without
considering the hysteresis effect may lead to unacceptable SOC
errors.
Fig. 4. Benefits and drawbacks of different electrical equivalent circuit models.
4
● Impedance look-up table method. There is a relationship between SOC was determined according to the difference between the
impedance and SOC. By applying a specific frequency of the current average cell and single cell. The results showed that the pack SOC
to the battery, several parameters related to SOC are identified by error is less than 2%. Whether the rated capacity is accurate directly
nonlinear fitting or parameter identification algorithm, then the affects the reliability of SOC estimation results. In Ref. [66], SOC and
impedance look-up table method is established [52]. The impedance capacity were jointly estimated by EKF, SOC error was further
parameters refer to internal ohmic resistance, polarization resis reduced. In Ref. [67], a robust estimate SOC method using EKF was
tance, polarization capacitance, inductance, constant phase element, proposed for the battery. The modeling error was regarded as a
etc [53]. If the impedance amplitude is small, these methods may constant bias state vector, which was used to correct SOC. The ac
cause significant prediction errors. Ref. [54] indicated battery aging curacy of the EKF depends not only on the linearization of the OCV
might affect the accuracy of SOC look-up table method. Besides, the function but also on model parameters [68]. Therefore, several
current, ambient temperature and other factors may lead to improved methods have been proposed to estimate battery SOC,
non-linear changes in impedance and SOC [55]. To ensure the reli which can be divided into two categories: (1) model improvement;
ability of SOC estimation, the effects of current ratio, aging, and (2) algorithm improvement. In Ref. [69], the
temperature on impedance need to be considered. thermal-electrochemical model was developed for the battery. The
model parameters are compensated by temperature and the SOC was
The primary defect of the look-up table method is that the battery estimated by EKF. In Ref. [70], an enhanced SOC estimation method
requires to be rested for a long time to ensure the stability of internal based on EKF was developed, which considered the effects of
electrochemistry so that the measured parameters can be relatively ac different currents, SOCs and hysteresis effects on the model param
curate. Besides, the reliability of SOC measurement dramatically de eters. In Ref. [71], the cubic Hermite interpolation was used to
pends on the accuracy of the SOC table. Therefore, the method is not correct the battery model, and the SOC was estimated by grey EKF.
suitable for the application of real-time and high accuracy of SOC esti The proposed method can overcome the influence of truncation
mation, such as aviation, aerospace, military industry, etc. error. In Ref. [72], a multi time scale estimation algorithm based on
EKF is proposed, which is applied to the joint estimation of SOC and
3.1.1.2. -hour integral method. Compared with the above methods, the SOH of battery. In order to reduce the error of SOC estimation, the
ampere-hour integral method is more straightforward, where the SOC of filter with better performance than EKF is used to estimate SOC.
the battery is calculated by current integration [56]. ● Adaptive Extended Kalman Filter (AEKF). In AEKF, the covariance
Although this method is simple, it has apparent shortcomings. Due to of process noise and observation noise are adaptive [73]. The
the open-loop calculation, the sensor error will be accumulated, which adaptive covariance of process noise and measurement noise makes
may generate more massive SOC error. Besides, aging and temperature AEKF effectively avoid the divergence or bias of the algorithm [74].
can cause deviations in rated capacity and Coulomb efficiency, which In Ref. [75], an on-line OCV estimation method was established by
affects the accuracy of SOC calculation. Moreover, the initial SOC is using AEKF, and SOC was determined by looking up the OCV-SOC
determined by the look-up table method so that any initial error may run table. In Ref. [76], the relationship between SOC and the chemical
through the whole SOC calculation process. Therefore, this method is composition of different types of batteries was analyzed. Then based
usually combined with the model-based method or data-driven method on a multi-parameter closed-loop feedback system, the accurate
to enhance its robustness. estimation of SOC was estimated by AEKF. The results showed that
the maximum SOC error is within 3%. Due to the aging of the battery,
3.1.1.3. Filter-based method. Typically, the filter-based method can be the accuracy of SOC estimation is challenged. A new method to es
roughly grouped into two categories: the Gaussian process-based filter timate SOC of LIB based on AEKF was proposed in Ref. [77]. A simple
approaches (LKF, EKF, AEKF, SPKF, UKF, AUKF, CDKF, CKF) and the optimization algorithm is applied to update the battery aging model,
probability-based filter approaches (PF, UPF, CPF). and the SOC with different aging batteries was estimated by AEKF.
The results showed that the SOC error is less than 4%. Due to the
● Liner Kalman Filter (LKF). For the past few years, Kalman Filter fractional-order model has a better description of the battery
(KF) is widely used for battery state estimation [57]. It can be seen as behavior, the battery SOC was AEKF estimated by AEKF based on the
a recursive process that involves two steps: (1) predicting the system fractional-order model [78]. In order to solve the problem of inter
state and output, (2) updating the system state based on output er ference of current measurement offset to SOC estimation, a joint
rors [58]. Because the OCV function is nonlinear, KF cannot be used estimation method of SOC and capacity based on AEKF multi time
directly. In Ref. [59], a method of SOC estimation of LIB using a LKF scale framework is proposed in Ref. [79], which effectively improves
based on local linearization was proposed. The OCV function is the robustness and accuracy of SOC estimation.
piecewise linearized, which makes LKF suitable for SOC estimation. ● Sigma-Point Kalman Filter (SPKF). For EKF, the linearization of
Chen et al. [60] proposed a joint estimation method of SOC and SOF OCV-SOC function is expanded near the prior mean, and the
based on LKF. Wei et al. [61] put forward a joint estimation method nonlinear part is ignored, which leads to obvious SOC estimation
of SOC and capacity. The SOC and capacity are estimated by LKF and error. Therefore, it is difficult for EKF to achieve satisfactory per
the recursive least squares, respectively. In recent years, a variety of formance in applications requiring high SOC estimation accuracy. In
KF variants are used for battery SOC estimation. Ref. [80], the SOC estimation methods based on SPKF ware pro
● Extended Kalman Filter (EKF). The principle of EKF is to linearize posed. The results indicated that SPKF could obtain higher SOC
the nonlinear system and perform Kalman filtering, so EKF is a estimation accuracy than EKF. In Ref. [81], a SOC estimation method
suboptimal filter [62]. The EKF expands the nonlinear OCV function based on an electrochemical model was proposed, which uses an
with partial derivatives based on the linearization principle of adaptive square root sigma point Kalman filter (ASRSPKF) with
nonlinear functions [63]. Aiming at the problem that the model equality constraints. The results show that ASRSPKF has outstanding
parameters are easily changed caused by the nonlinear behavior of performance. Compared with AEKF, its accuracy is improved by
the battery, the SOC estimation method based on a reduced-order 30%, and its convergence time is shortened by 88%.
battery model and EKF was proposed in Ref. [64]. Experimental re ● Unscented Kalman Filter (UKF). UKF is developed from the
sults showed that SOC errors are within 2%. In Ref. [65], a SOC traceless transformation and standard KF. By traceless trans
estimation method based on a dual-time scale EKF was developed for formation, nonlinear system equations can be applied to the standard
the battery pack. First, the average SOC was estimated, then the cell Kalman filter under linear assumption [82]. UKF does not ignore
higher-order terms due to linearization, so it has high estimation
5
accuracy [83]. The parameters of the battery model are sensitive to covariance information by the Cholesky decomposition, which im
temperature. In Ref. [84], model parameters were compensated by proves the performance of CKF in SOC estimation [99].
temperature, then SOC was estimated by UKF. To improve the per
formance of traditional UKF, several improved UKF algorithms were The Kalman filter and its distortion are limited by the Gaussian
developed. In Ref. [85], the battery model was trained by the radial distribution condition, so it is difficult to obtain a satisfactory filtering
basis function (RBF) network, the SOC was estimated by SRUKF. The effect for non-Gaussian distribution. At present, the more popular way to
results showed that SRUKF has higher SOC estimation accuracy than solve the generic filter is to use the particle filter (PF) [100,101].
EKF. To reduce the calculation requirements of unscented trans Nowadays, the research of PF has made gratifying progress, and many
formation in UKF, the square root unscented Kalman filter using deformations have been proposed used in battery state estimation.
spherical transform (Sqrt-UKFST) was developed to estimate battery
SOC in Ref. [86]. Compared with EKF, RMSE and maximum error ● Particle Filter (PF). The core idea of PF is to generate a set of
increased by 37% and 44%, respectively. The calculation require discrete sampling points in the state space according to the empirical
ment is reduced by 32% than UKF. To improve the robustness of distribution of the system state vector, then adjust the position and
UKF, the fuzzy inference system was applied to optimize UKF [87]. state of particles according to the observed values. Finally, the
The improved UKF algorithm is more robust, and SOC error under optimal particle state is estimated by adjusting the particle sets [102,
the UDDS condition is within 1.76%. 103]. A method for estimating SOC and residual discharge time
● Adaptive Unscented Kalman Filter (AUKF). AUKF is an upgrade of (RDT) of battery based on PF was developed in Ref. [104]. In
UKF, which can automatically adjust the noise covariance in the SOC Ref. [105], the PF was applied to estimate SOC and SOH. In
estimation [88]. The AUKF was employed for battery SOC estimation Ref. [106], the performance of PF and EKF in SOC estimation was
in Ref. [89]. The recursive least squares method identified model compared. The results showed that PF is better than EKF. To reduce
parameters on-line and updated the AUKF state model in real-time. SOC estimation error, a SOC estimation method based on dual PF was
The results showed that the joint estimation algorithm could proposed in Ref. [107]. The numerical results indicated that the
reduce the SOC estimation error. Mara et al. [90] have compared the proposed method has higher accuracy then EKF and UKF.
performance of AUKF, AEKF, UKF, and EKF in SOC estimation based ● Unscented Particle Filter (UPF). The UPF is an improved form of
on a 2nd order RC model. The results indicated that AUKF has the PF. Its main idea is to use UKF to improve the sampling process of PF.
best performance. The absolute average error was only 0.028%. To The UKF calculates the mean and variance for each particle, then PF
calculate the error covariance adaptively, an improved AUKF based resamples the particle. UKF uses the function of approximate pos
on Sage-Husa maximum posterior estimation was proposed for SOC terior filter density to calculate the mean and variance of particles.
estimation. In Ref. [91]. The experimental results showed that the The latest observation information is included in the calculation
accuracy of SOC estimation is higher than that of UKF and EKF. results, so the PF sampling effect is improved [108]. Due to its su
● Central Difference Kalman Filter (CDKF). The CDKF assumes that perior performance, UPF has been widely used in SOC estimation. In
the state variables of the system obey Gaussian distribution. For Ref. [109], the author compared the algorithm performance of EKF,
random Gaussian variables with known mean and covariance, CDKF UKF and UPF in SOC estimation. The comparison results showed that
estimates their mean and covariance after any nonlinear trans the numerical indexes of UPF are optimal. In Ref. [110], a prediction
formation. CDKF uses the Sterling interpolation formula to expand framework of SOC and RDT based on UPF was proposed for the
the nonlinear model according to the central difference, which does battery. The UPF improved the accuracy of SOC estimation and
not need to calculate the partial derivative of the function [92]. The ensured the reliability of RDT prediction. For UPF, if the number of
accuracy of EKF and CDKF in estimating SOC was compared in particles is too large, the convergence speed of the algorithm will
Ref. [93]. The results showed that CDKF has a higher performance in slow down. To overcome this issue, a hybrid UPF algorithm was
estimating SOC. The SOC absolute maximum error is less than 6%. To proposed in Ref. [111], which uses the normalized least mean square
avoid the high-order Taylor series expansion in EKF calculation, an method to adapt the search range, so as to improve the efficiency of
improved CDKF algorithm using second-order difference trans UPF. The results showed that the prediction SOC estimation error is
formation to generate sigma points was proposed in Ref. [94]. The within 2%, and the improved algorithm has better performance than
verification results showed that the SOC error is within 2%, which is the original algorithm.
better than EKF and UKF. ● Cubature Particle Filter (CPF). CPF directly calculates the mean
● Cubature Kalman Filter (CKF). The CKF was proposed by Canadian and variance of nonlinear random function by volume method and
scholars Araararatnam and Haykin in 2009 [95]. Based on the generates the suggested density function to get the weighted parti
third-order spherical radial volume criterion, the mean value of the cles. Then, by calculating the mean value of the particles, the mini
nonlinear system state is approximated by CKF using a series of mum mean square error estimation of the system state is obtained
volume points, which can effectively solve the problem of nonlinear [112]. CPF algorithm uses the latest measurement information when
system state estimation. The most important two steps are to trans generating particles, which improves the approximation degree of
form the integral form into the spherical radial integral form and the the posterior probability of system state [113]. The complexity,
third-order spherical radial criterion. In Ref. [96], the performance precision and robustness of three SOC estimation methods based on
of CKF and EKF in SOC estimation was compared. The experimental extended particle filter (EPF), CPF and UPF were compared in
results show that the maximum absolute error of CKF-based SOC Ref. [114]. The experimental results indicated that the EPF algo
estimation of CKF is 1.78% lower than that of EKF. However, the rithm is efficient, but SOC estimation error is high. On the contrary,
calculation time of CKF is 8.8 times that of EKF. In Ref. [97], a SOC CPF and UPF can improve the accuracy of SOC estimation, but
estimation method based on adaptive CKF (ACKF) was developed. require more running time. In Ref. [115], an adaptive weighted cubic
The results showed that the robustness and convergence rate of ACKF particle filter (AWCPF) was proposed to estimate SOC. Aiming at the
are better than that of CKF. In Ref. [98], the improved cubature problem that the CKF algorithm is sensitive to noise, the proposed
Kalman filter (ICKF) based on singular value decomposition (SVD) algorithm adaptively adjusts the sigma point weight according to the
and Gauss-Newton iterative technique was used to estimate battery state and measurement residual vector iteratively. The algorithm
SOC, which reduces the SOC estimation error. To ensure the was used to verify different operating conditions with various initial
non-negativity of the covariance matrix and avoid filter divergence, SOC. The results showed that AWCPF has high estimation accuracy,
the square root cubature Kalman filter (SRCKF) calculates the state strong robustness, fast convergence speed, and the maximum SOC
estimation error is within 1%.
6
3.1.1.4. Observer-based method. The state observer can obtain the esti ● Other Observers. Wei et al. [133] developed a recursive total least
mated values of the state variables based on the measured values of the squares (RTLS) based observer to address the biased identification of
external variables of the system. In order to realize state feedback or model parameters resulting from the unexpected sensing of noises. A
other needs for control systems, Luenberg [116] proposed the concept RTLS based on Rayleigh quotient (RQ) minimization was employed
and construction method of state observer. The appearance of the state to minimize the influence of current and voltage sensing noises. A
observer not only provides a practical possibility for the realization of Frisch scheme based bias compensating recursive least squares
state feedback technology but also has been applied in many aspects of (FBCRLS) based observer was proposed in Ref. [134]. The FBCRLS
control engineering. In recent years, the observer-based methods such as method was utilized to estimate online the noise variances and the
the Luenberger observer (LO), the sliding mode observer (SMO), the unbiased model parameter. Then the FBCRLS was integrated with a
proportional-integral observer (PIO), and the H-infinity/H∞ observer SOC observer in a closed-loop feedback mechanism, namely FBCRLS
(HIO) have been used extensively for battery state estimation. based observer. Ma et al. [135] proposed an input-to-state stability
(ISS) theory based nonlinear observer for the SOC estimation. Two
● Luenberger Observer (LO). The LO is widely used in linear, nonlinear robust observers based on a reduced-order electrochemical
nonlinear and time-varying systems [117]. Hu et al. [118] proposed model were presented in Ref. [136]. Both observers contained a
an adaptive Luenberger observer (ALO) based method for online SOC Luenberger term for nominal error convergence and a variable
estimation of a battery pack. The observer gain was adaptively structure term for the improvement of robustness of model un
adjusted through a stochastic gradient approach. Wang et al. [119] certainties. Li et al. [137] proposed a discrete-time nonlinear
developed a LO for SOC estimation based on a nonlinear fractional observer for battery SOC estimation. Compared with the EKF and the
battery model. The global asymptotic stability can be ensured discrete-time SMO algorithms, the computation cost was reduced
through Lyapunov’s direct method. and the estimation accuracy and convergence speed were improved.
● Sliding mode observer (SMO). The SMO is developed from the Xia et al. [138] applied a special nonlinear observer to verify the
sliding mode controller (SMC) and maintains robust tracking per improved parameter identification method based on polarization
formance under the condition of model uncertainties and environ depth.
mental interference. A switching gain is designed to ensure stability
and convergence. To eliminate the chattering phenomenon, a 3.1.1.5. Data-driven based method. The data-driven based methods
second-order discrete-time SMO was developed in Ref. [120]. In consider the battery as a black box and learn the internal dynamics
Refs. [121,122], an adaptive gain SMO (AGSMO) was proposed to through large amounts of measurable input and output data. The
compensate model errors and minimize chattering levels. A grey commonly used data-driven based methods for SOC estimation include
prediction-based fuzzy SMO (GPFSMO) was proposed in Ref. [123] neural network (NN), fuzzy logic, genetic algorithm (GA), support vec
to both reduce chattering and solve the over-estimation problem. tor machine (SVM), and so on.
The value of battery voltage was predicted through grey prediction
and the gains on SMO were tuned according to the prediction error ● Neural Network (NN) method. The NN method has an excellent
through fuzzy inference system. Chen et al. [124] proposed a robust ability to form a non-linear map to demonstrate a complex nonlinear
SMO (RSMO) for SOC estimation. The switching gain of RSMO was model. Dang et al. [139] proposed an OCV-based SOC estimation
adjusted according to the upper bound of system uncertainty that is method on the basis of the dual NN fusion battery model. The linear
learned by a radial basis function neural network (RBFNN), and NN battery model was used to identify parameters of the first-order
achieved the convergence of the SOC estimation errors. A or second-order electrochemical model, and the second
fractional-order SMO based SOC estimation method was designed in back-propagation NN (BPNN) was utilized to capture the relation
Ref. [125]. Huang et al. [126] proposed a novel super-twisting SMO ship between OCV and SOC. Sun et al. [140] developed a Radial Basis
(STSMO) based SOC estimation method. The results indicated the Function Neural Network (RBFNN) based uncertainty quantification
STSMO can achieve faster convergence speed and higher robustness algorithm to construct response surface approximate model (RSAM)
with less computation cost. of model bias function to estimate the SOC of multi-cell battery pack.
● PI Observer (PIO). The PIO is an efficient method to estimate the Tong et al. [141] established a load-classifying NN model to estimate
state of systems with unknown input disturbance. A PIO based SOC the SOC. The structure of the model and the post-processing improve
estimation method based on a simple RC battery model was proposed the suppression of over fitting. Chemali et al. [142] developed a long
in Ref. [127]. Tang et al. [128] developed a dual-circuit based short-term memory recurrent neural network (LSTM-RNN) based
observer. The parameters-normalized PIO was designed to deal with SOC estimation for Li-ion batteries. A BPNN based SOC estimation
the capacity error and initial error, and the current integrator was to strategy was introduced in Ref. [143]. In addition, principal
restrain the influence of the drifting current. The experiment results component analysis (PCA) and particle swarm optimization (PSO)
showed that without matrix operation, this method had lower were utilized to improve estimation accuracy and robustness. A
computation complexity but high accuracy, even when the initial novel method based on deep feed-forward NN was used for battery
SOC was unknown. SOC estimation in Ref. [144] and the battery measurements can be
● H-infinity/H∞ Observer (HIO). The HIO can ensure the robustness directly mapped to SOC. Chen et al. [145] proposed a RBFNN based
of the inaccurate initial system state and unknown disturbance from nonlinear observer using an inclusive equivalent circuit model to
the inaccurate or unknown statistical characteristics of modeling and estimate SOC. An improved nonlinear autoregressive with exogenous
measurement errors. To accommodate the mode uncertainties from input (NARX)-based NN (NARXNN) algorithm was developed to es
current, temperature, and aging, a methodology combined with an timate the SOC of batteries [146]. The lighting search algorithm
HIO and a hysteresis model was proposed in Ref. [129]. Zhu et al. (LSA) was used to find the best value of input delays, feedback de
[130,131] developed an HIO with dynamic gain for the SOC esti lays, and hidden layer neurons. Xia et al. [147] proposed a
mation of a battery pack, which can reduce the adverse influence of multi-hidden-layer wavelet NN (WNN) model optimized by
the non-Gaussian model and measurement errors. In addition, the Levenberg-Marquardt (L-M) algorithm. PSO algorithm was
observer design criterion was formed as the linear matrix inequality employed to optimized WNN for SOC estimation. Yang et al. [148]
(LMI) for easy computation. Liu et al. [132] developed an HIO using developed a RNN with gated recurrent units to estimate the battery
a switched battery model to estimate the Electromotive force (EMF) SOC. A stacked LSTM network was proposed in Ref. [149] to model
for SOC. The switched model considered the relaxation effect and the the dynamics of lithium iron phosphate batteries and estimate the
relationship between EMF and SOC.
7
SOC. The proposed method presented quick convergence to the true EKF-based method depends on the linearization error. Thus, this is un
SOC even if the initial SOCs were inaccurate. suitable for high order nonlinear systems. To avoid the complicated
● Fuzzy Logic method. A fuzzy NN (FNN) based battery model was Jacobian matrix and Gaussian noise computation, the SPKF- and UKF-
established to capture the battery nonlinear dynamics [150,151]. Lee based methods are developed for SOC estimation, but more computing
et al. [152] proposed a learning system that consists of learning resources are required. Adaptive methods, such as AEKF and AUKF, are
controllers, FNNs, and cerebellar-model-articulation-controller net further developed to enhance robustness against the measurement
works to estimate the SOC. Li et al. [153] developed a noises and processing noises. To deal with the non-Gaussian noises, PF-
simple-structure merge FNN for SOC estimation, where the soft based method is proposed and can achieve high accuracy. Furthermore,
computing approach combined the FNNs based on B-spline mem the UPF- and CPF-based methods are designed to improve particle
bership function (BMFs) and a reduced-form genetic algorithm sampling process and state probability approximation, but the conver
(RGA). gence rate remains a challenge. The observer-based methods can also
● Genetic Algorithm (GA). The GA is usually used to identify the obtain satisfactory estimation accuracy against sensor or initial state
battery model parameters for further SOC estimation [154]. Chen errors. However, the appropriate gain determination is still a critical
et al. [155] developed a novel SOC estimation method on the basis of issue. Unlike the model-based methods, the data-driven based methods
the grey model (GM) and genetic algorithms. The utilization of a are insensitive to the model accuracy and environmental conditions. The
genetic algorithm introduced higher accuracy and repeatability. main drawbacks are that these methods usually need high computa
● Support Vector Machine (SVM). The SVMs are a cluster of related tional cost and long processing time, and the accuracy depends on the
supervised learning methods that can universally approximate any trained data to a great extent.
multivariate function to any level of accuracy [156]. An optimized
SVM for regression based SOC estimation method was proposed in 3.1.2. Review of state-of-energy estimation approaches
Ref. [157]. The results verified this method was simpler and more The state-of-energy (SOE) was first proposed by Mamadou et al.
accurate than that based on artificial neural networks (ANNs). Meng [160]. The SOE estimation approaches are similar to that of SOC esti
et al. [158] combined the adaptive unscented Kalman filters (AUKF) mation. The most commonly used approaches are data-driven based
and least-square support vector machines (LSSVM) to estimate bat approaches and adaptive filter-based approaches. Liu et al. [161] used
tery SOC. The battery model can be accurately established and the SOE to estimate the remaining energy and considered the energy loss
updated even with limited training samples. Deng et al. [159] of internal resistance, chemical reaction, and OCV reduction. They also
implemented a least squares SVM prediction algorithm to predict the deeply explored the impacts of temperature and C-rate on SOE estima
battery available capacity. tion and improved the accuracy of the estimation. Wang et al. [162]
proposed a joint estimator of SOC and SOE based on particle filtering.
The comparison of the above mentioned SOC estimation approaches They performed experiments on IFP1865140-type batteries under dy
in terms of computation accuracy, application scenarios, benefits and namic temperature and verified the robustness and accuracy of the
drawbacks, is summarized in Fig. 5. The look-up table method is simple method. He et al. [163] proposed a battery SOE estimator based on a
and needs to establish the nonlinear relationship between SOC and OCV Gaussian model. The results showed that the estimated SOE can guar
or impedance offline. Its main disadvantage is that it can not meet the antee high accuracy. Zhang et al. [164] proposed a joint estimator for
real-time requirements and is sensitive to the accuracy of sensors. The SOE and SOP by using AUKF. The estimated error of battery SOE is less
ampere-hour integral method is simple and needs low computational than 2%. Dong et al. [41] established a battery SOE estimator using PF.
cost. However, this approach is open-loop and the accuracy mainly re The battery state-space model was built based on a wavelet neural
lies on the initial SOC and the sensor accuracy. Thus, this method network. The superiority of this method was verified on LiFePO4 bat
combined with other methods like OCV look-up table or model-based teries. Zhang et al. [165] proposed an adaptive H-infinity observer to
methods is a promising approach for better SOC estimation accuracy. estimate SOC and SOE of Li-ion batteries. The proposed method was
The model-based methods, including filter-based and observer-based verified to be more accurate than the EKF method. Zheng et al. [166]
methods, can achieve high estimation accuracy and robust to noise. comprehensively analyzed the correlations between ambient tempera
Nevertheless, the accuracy of the model-based methods usually depends ture, C-rate, battery aging degree, and the maximum available energy of
on the model precision and more computing resources are required due the battery. Dong et al. [167] proposed a dual filtering algorithm based
to algorithm complexity. The KF family based methods are widely used on EKF and PF, and established an online SOE estimator. The estimated
in SOC estimation. The LKF-based method is only suitable for the linear error was around 4%. Lin et al. [168] proposed a multi-model proba
systems. To overcome this drawback, the EKF-based method applies the bilistic estimation method, which combined the estimation results of
linearization approximation to the nonlinear systems. The accuracy of different models according to the weights determined by Bayes theorem.
Fig. 5. The comparison of different SOC estimation approaches.
8
3.2. Power capability prediction will be considered, such as physics-based battery models and associate
state constraints.
The prediction of power capability is also crucial in battery man
agement which shows users how much power is available in the im 3.2.3. Data-driven approach
mediate future. The power capability is the rate at which energy can be In the data-driven approach, the battery is regarded as a black box,
moved from the battery pack to the loads without exceeding cell or and the reaction mechanism and characteristics inside the battery are
electronics design limits and is an instantaneous quantity. Generally, the not considered. The SOP is taken as output of a model by means of data
state-of-power (SOP) or state-of-function (SOF) is used to evaluate the analysis and machine learning, and influence factors such as voltage,
power capability of a battery. The SOP can be defined as the percentage temperature, and SOC are taken as inputs. Fleischer et al. [182] pre
of peak power relative to rated power, where the peak power is the sented a method to estimate SOP by an adaptive neuro-fuzzy inference
maximum continuous power over a short period of time that does not system (ANFIS).
exceed the thresholds [169]. According to the available literature, the
common SOP calculation methods can be divided into three types:
experimental testing approach, multiple constraints estimation 3.3. Lifespan and health prognoses
approach, and data-driven approach.
The state-of-health (SOH) is a critical index for lifespan and health
3.2.1. Experimental testing approach prognoses, which can reflect the current life condition of batteries [183].
The measurement approach for battery power capability evaluation The definition of SOH can be expressed as shown in Eq. (1).
based on the hybrid pulse power characterization (HPPC) test is pro SOH ¼ Qmax =Qnew (1)
posed in Ref. [170]. The experimental testing approach is the most
commonly used approach for static peak power measurement. It can In Eq. (1), Qmax is the maximum available capacity of the battery,
estimate the instantaneous available power. However, it needs a static Qnew is the available capacity of a new battery. The SOH/RUL estimation
test environment resulting that it is impractical in practical applications. approaches can be divided into the following groups: measurement &
analysis approach, Bayesian-based estimation approach, empirical
3.2.2. Multiple constraints estimation approach fitting approach, and machine learning-based approach.
In order to overcome the drawback of the experimental testing
approach, Plett et al. [171] proposed a voltage-constraint method to 3.3.1. Measurement & analysis approach
improve the HPPC method which can estimate the continuous useable
power. Feng et al. [172] proposed a SOP prediction algorithm based on 3.3.1.1. Direct measurement approach. The battery internal resistance
an equivalent circuit model, considering both the voltage and current and available capacity are critical parameters for the battery SOH
limits. However, it may cause over-charge or over-discharge without assessment. The Coulomb counting method is a useful method for ca
SOC as a constraint when the batteries are nearly saturated or hungry. pacity estimation. In Ref. [184], the Coulomb counting method
Dong et al. [173] analyzed the influenced factors of SOP based on sta employed to estimate the SOH is evaluated by the maximum releasable
tistical data. Moreover, they proposed a joint estimator for SOC and SOF capacity. Kim et al. [185] developed an enhanced Coulomb counting
of LIB based on KF. The SOF in this paper indicated the power capability algorithm that can get robust and accurate SOC and SOH estimation. The
of the battery, which was consistent with the content expressed by the internal resistance is an important indicator of battery degradation.
SOP. The methods use only one constraint to estimate the SOP of the Blanke et al. [186] presented an overview of methods used to estimate
battery would give over-optimistic results, due to the peak current the SOH based on impedance measurements. The EIS technique is
would not be allowed to be used under a large-scale SOC [174]. Hence, a employed to determine the SOH through the change of the impedance
multi-constraints (voltage, current, power, SOC, etc.) SOP prediction spectrum at a particular peak [187].
algorithm was proposed to obtain a more precise result in Ref. [175].
The advantages and disadvantages of the single-constraint methods 3.3.1.2. Indirect analysis approach. In Refs. [188], the indirect analysis
were deeply analyzed in this paper. Furthermore, the results indicated of the battery charging and discharging curves are employed for battery
that the multi-constraints method could overcome the drawbacks of the SOH estimation. In Ref. [189], incremental capacity analysis method is
single-constraint methods and improve the precision and robustness of used to obtain the battery aging level. The support vector regression is
their model. Wei et al. [176] established an adaptive peak power esti employed to extract the aging level and the results show that this
mator for vanadium redox battery by combining voltage, SOC and cur method can predict the SOH within 1% error bound. Merla et al. [190]
rent design constraints. Experiments verified that the method has good proposed a novel technique called differential thermal voltammetry
performance in modeling accuracy, convergence, and computational (DTV) to determine the SOH of batteries. The peak parameters of DTV
complexity. Zhang et al. [177] proposed an adaptive UKF joint estimator which represent the particular chemical reaction can be used to explore
to jointly estimate the SOE and SOP of the battery. The effectiveness of the aging process [191,192]. In Ref. [193], the differential voltage (DV)
the method at different temperatures and different aging degrees was curve is used to quantify the aging mechanisms. Ref. [194] Presented the
verified on lithium iron phosphate batteries. Sun et al. [178] proposed a center least squares method to obtain the DV curve. The inflection point
joint estimator of AEKF algorithm to jointly estimate battery SOC and location and the transformation parameters can be used in battery SOH
SOP. Wang et al. [179] estimated the power and energy of lithium-ion estimation. The differential mechanical parameter (DMP) analysis is a
batteries and ultra-capacitor hybrid systems. They carried out a complementary approach to the existing SOH estimation techniques
detailed analysis of power capacity and maximum charge/discharge [195,196]. Cannarella et al. [197] developed a DMP-based approach to
energy prediction under multiple constraints. Experiments and discus estimate SOH by using the stress measurement.
sions on the prediction of charge and discharge energy are conducted on
different timescales. Tang et al. [180] proposed a high-fidelity migration 3.3.2. Bayesian-based estimation approach
model for SOC and SOP estimation. An efficient numerical search al
gorithm is used to calculate the value of the SOP, and it is constrained by 3.3.2.1. Kalman filters family. The KF can be applied to a linear system
multiple conditions such as current, SOC, and voltage. Zou et al. [181] with Gaussian noise [198,199]. In Ref. [200], the KF is used to predict
for the first time formulates battery power prediction and management the capacity of batteries using a two-phase service model. Wassiliadis
as an economic model predictive control. The algorithm will be et al. [201] used two EKF to obtain the battery states and model pa
extended in this application for battery management where more factors rameters synchronously. However, this method will deviate from the
9
real value at the end of battery life. To solve this problem, Xue et al. and the complex internal electrochemical process, Dai et al. [222] pro
[202] proposed an AUKF realize the multi-step prediction of SOH and posed a novel SOH estimation method by using prior knowledge-based
RUL by updating the noise covariance. In Refs. [203,204], the EKF is neural network (PKNN) and Markov chain. Fen224g et al. [223]
compared with the knowledge-based algorithm. The result shows that developed a predictive diagnosis method by comparing partial charging
the EKF and the knowledge-based algorithm can obtain similar results. curves with the stored SVMs. The Gaussian process regression (GPR) is a
type of Bayesian non-parametric machine learning method, which can
3.3.2.2. Particle filters family. The PF also can be used in battery life handle uncertainty in a complex model. Ref. [224] capture the mapping
span and health prognoses, which is suitable for nonlinear systems with among capacity, temperature, and SOC by using GPR with an automatic
non-Gaussian noise [205–207]. To deal with the aging fluctuations in relevance determination structure. Ref. [225] consider the electro
the uncontrolled environments, Tang et al. [206] proposed a chemical knowledge to the understanding of the covariance functions
model-oriented gradient-correction particle filter (GC-PF) based method within the GPR. In Ref. [226] the long short term memory (LSTM) model
to predict the aging process of Lithium-ion batteries. Miao et al. [208] is applied to predict the residual of GPR. Richardson and Howey et al.
used the UPF to estimate the remaining useful life (RUL) of batteries [227] proposed a GPR method for forecasting battery SOH, which
with an error of less than 5%. In Ref. [209], the Gauss Hermite PF showed good performance than other data-driven approaches.
technique is proposed to predict the battery capacity fading based on a The key benefits, drawbacks, accuracy, and application scenarios of
state projection scheme method. Zhang et al. [210] proposed a RUL different approaches are summarized in Fig. 6. The measurement and
prediction method that combined the PF with the exponential degra analysis approach is the simplest and direct method, which need to
dation model due to the nonlinear and non-Gaussian fading phenome calculate and analyze the measured external characteristics of batteries.
non. The results show that this method has better prediction accuracy This method is open-loop, so the accuracy is affected by noise and
than the regularized PF according to the accuracy index. Wang et al. environmental disturbance. The empirical fitting approaches need to
[211] developed a prognostic technology using the spherical cubature establish the aging model bade on experience and electrochemical
PF (SCPF) based on the state-space model to predict the RUL of batteries. knowledge. A large number of battery aging test data are needed to fit
For PF method, the importance function and the degradation of the the model. Therefore, specific models are often only effective for the
sampling particles influenced the estimation accuracy [212]. Therefore specific type of battery and the specific aging condition. The Bayesian-
Zhang et al. [213] proposed the UPF to predict the RUL based on linear based approaches are developed from the previous methods. These
optimizing combination resampling (LOCR), which uses UKF to generate methods estimate the key parameters to evaluate the battery health state
a distribution as the importance function and use LOCR to improve the by the closed-up filtering algorithm, which are robust to the modeling
diversity of sampling particles. and measurement error. The machine learning-based approaches are the
black box model, which is intelligent and flexible, but the estimation and
3.3.3. Empirical fitting approach prediction result is sensitive to the training data set.
The empirical fitting method is usually based on battery models that
relate to the degradation of one or more performance parameters. In 3.4. Other indexes in battery management system
Ref. [214], a memory-less degradation model is developed, which al
lows the degradation rate to depend on both the stress level and SOH. 3.4.1. Review of remaining discharge time prediction approaches
The results showed that this model can describe a broader range of The remaining discharge time (RDT) indicates the continuous oper
battery fading behavior. Wang et al. [215] established a generalized ation time until the voltage of the battery reaches its lower threshold,
battery life model considered the using time, C-rate, and temperature. In which is helpful to solve the range anxiety issues. Although the SOC and
Ref. [216], the calendar aging test and cycle aging test are conduct to SOE can reflect the residual capacity and energy of the battery, they
analyze different aging factors. An aging model is developed with cannot give the driver an exact endurance time directly. The crucial
impedance-based model that based on the electric-thermal kinetics to issues for RDT prediction are the modeling of the battery system and the
obtain the mathematical description of degradation. Smiley et al. [217] forecast of future information such as discharge current and terminal
proposed a method to estimate the parameters of the aging model and voltage profile [104]. The Markov transition probability model can be
select the most representative model from a set of pre-aged models using used for battery current discharge profile prediction. In Ref. [228], Pola
the interacting multiple-model KF and the Viterbi algorithms. Zhang et al. employed a PF-based method considering the battery current
et al. [218] developed a capacity loss model which is coupled with the transition probabilities for a two-state Markov chain. Wang et al. [35]
aging mechanisms of solid electrolyte interface layer growth and active established fractional-order models and proposed a Bayes Monte Carlo
material loss. The recursive least squares method is applied to estimate estimator for RDT prediction with future load trajectory information
the model parameters. The result shows that the estimated capacity loss obtained by a Markov probability model. For known past dynamics of
has less than 1% error. Zou et al. [72,181] proposed a multi-time-scale input currents, a discrete wavelet transform technique was employed to
estimation algorithm to predict the SOH and formulate an economic extract future battery currents [229]. The recursive method also can be
predictive model for battery health management. The algorithm will be used to predict a future constant-current by introducing forgetting fac
extended in this application for battery management, where more fac tors. Quin
~ ones et al. [230] employed the Lambert function together with
tors will be considered, such as physics-based battery models and an electrochemical reduced-order model for RDT prediction. Wang et al.
associated state constraints. [110] proposed a prediction framework for SOC and RDT based on an
equivalent circuit model considering hysteresis. In Ref. [231], a
3.3.4. Machine learning-based approach Dirichlet process mixture model was established to identify the voltage
With the development of big data and artificial intelligence industry, trajectory which can be utilized for RDT prognostic. Considering the
machine learning methods have been applied to predict battery SOH and dynamic uncertainties in the RDT prediction process, Zhang et al. [232]
RUL [219]. Chang et al. [220] developed a RUL prediction method quantified the uncertainty of state estimation and future loads with the
composed of relevance vector machine (RVM), UKF and complete probabilistic method.
ensemble empirical mode decomposition (CEEMD). The RVM is used to
correct the prognostic results obtained by UKF. For the estimation of 3.4.2. Review of state-of-temperature estimation approaches
SOH and RUL, the SVM is a well-known machine learning method. Thermal management is a key part of the BMS, and accurate state-of-
Nuhic et al. [221] proposed an on-board SOH and RUL monitoring temperature (SOT) estimation is a high priority of thermal management.
method using SVM. In order to handle the uncertain external conditions High temperatures would adversely affect the performance and lifespan
of batteries, resulting in thermal runaway, e.g., fire and explosion [233].
10
Fig. 6. Key benefits and drawbacks of different SOH/RUL estimation approaches.
The surface temperature of battery can be measured easily with some of injecting excitation signal for EIS test, Wang et al. [251] proposed a
conventional measurement methods such as thermocouples. However, fast impedance calculation method based on wavelet transform. In
the internal cell temperature during transients differs significantly (in battery pack application, Beelen et al. [252] took crosstalk interference
the experimental study of Ref. [234] more than 20%) from temperatures in surrounding cells and influence of currents flowing through the pack
measured by external sensors located on the cell surface [235]. There into account. The robust estimation of battery system temperature under
fore, monitoring the internal temperature of battery is more crucial and sparse sensing and uncertainty was studied in Ref. [253]. By using in
challenging for thermal management. The SOT estimation approaches dicators of electrochemical impedance, internal temperature can be
can be divided into the following groups: direct measurement approach, mapped uniquely. But there are temperature gradients in battery, and
model-based estimation approach, electrochemical impedance-based large difference exists between the maximum temperature and the mean
approach, and data-driven approach. temperature. Richardson et al. [254] presented a promising new
approach for estimating the internal temperature distribution in cylin
3.4.2.1. Direct measurement approach. Researchers have proposed drical cells by combining measured electrochemical impedance and
methods of battery internal temperature measurement by integrating thermal model.
temperature micro-sensors into the battery cell internal layers [236].
For battery internal temperature indication, the commonly used sensors 3.4.2.4. Data-driven approach. The data-driven approaches have been
are thermocouples [237] and resistance thermometers [238]. The used in estimating the internal temperature of the battery. In Ref. [255],
traditional temperature measurement methods and several recently a data-driven approach combining the RBF neural network and the EKF
introduced methods such as fiber Bragg-grating techniques are reviewed is proposed to estimate the internal temperature for lithium-ion battery
in Ref. [239]. Although battery internal temperature can be obtained thermal management.
directly by inserting sensors, this approach would bring additional
manufacturing costs and potential safety threats. 3.4.3. Review of state-of-balance estimation approaches
The state-of-balance (SOB) is a novel concept for the consistency
3.4.2.2. Model-based estimation approach. Numerical thermoelectric evaluation which is designed for the purpose of the balance state char
and thermal models are widely used in internal temperature estimation acterization and regulation. SOB characterized not only the cell capacity
[240]. In order to establish thermoelectric and thermal models such as balance, but also the balance state for the LIB pack. The factors of cell
lumped-parameter battery model [241] and distributed battery thermal voltage, pack voltage, temperature, maintenance current, and internal
model [242], the knowledge of heat generation, conduction, dissipation, resistance are considered to estimate SOB [256]. In Ref. [256], the SOB
balance, and thermal boundary conditions are indispensable. Basing on estimation was realized with the detailed deduction, in which a dual
the thermal model established, Kim et al. [243] and Dai et al. [244] filter consisting of the UKF was used in order to predict the balance state
applied the adaptive KF method for the estimation of the battery core of the battery pack. In order to characterize the cell-to-cell battery
temperature. Ref. [245] Presented an experimental method based on consistency, Wang et al. [257] proposed an iterative SOB evaluation
thermal conduction analysis to determine the core temperature of a method by conducting the improved variation coefficient calculation.
Li-ion cell during thermal runaway. The battery internal temperature The SOB and aging process correction were considered.
may be well estimated based on the established thermal or thermo
electric model and state estimation approach, but during actual opera 4. Challenges and outlooks
tion of electric vehicles, various complex situations should be taken into
account, which is quite different from lab environment. With the popularization of energy storage technology, the develop
ment of advanced battery management technology has attracted more
3.4.2.3. Electrochemical impedance-based approach. Without establish and more attention. The battery itself is a kind of complex electro
ing a thermal model, some other methods for battery temperature esti chemical system. It is difficult to accurately model the battery system,
mation have been proposed based on EIS measurements [246]. and estimate the battery states, which seriously influences the reliability
Impedance indicators obtained from EIS, such as phase shift, real part and effectiveness of the BMS. Therefore, we outline some research di
amplitude, and imaginary part amplitude, can map to temperature. For rections and trends in this field from the following aspects.
instance, the phase shift was utilized to estimate internal temperature in
Refs. [247]. Since the impedance is a composite function of SOC, SOH, 4.1. Advanced sensing
and temperature, the impedance indicators should be chosen at special
frequency where the indicators are not sensitive to SOC and SOH [248]. The inside of the battery is a closed space, so the internal multiple
Schindler et al. [249] and Barai et al. [250] studied the battery states and parameters are hard to be directly measured. Thus, many
impedance with different relaxation processes. Considering the impact researchers have studied the battery state estimation algorithms and
parameter identification methods. However, these indirect approaches
11
will inevitably bring estimation or identification errors. The more ac Song et al. [269] proposed a data-driven least squares support vector
curate way to obtain internal multiple states and parameters is to use machine and model-based UPF method for joint estimation of SOC and
advanced sensors to directly measure them. For example, fiber-optic SOH of lithium-ion batteries. Feng et al. [270] designed an accurate joint
sensing technology has been introduced to monitor the internal bat estimator of SOC and SOT. The temperature range of their experiment is
tery states. Fiber optic sensors can be embedded into the Li-ion battery from 10 � C to 40 � C, the temperature change rate is as high as 10 � C. In
pouch cells to monitor the internal strain and temperature [258–260], this verification process, their method still showed good performance.
and then these signals can be used for high-accuracy cell state estima There are other joint state estimates, such as SOC and SOP, and SOC and
tion. A buckled membrane sensor has been developed for in situ, RUL. However, researches on the joint estimation of three or more types
simultaneous characterization of stress and microstructure evolutions in still need to be deepened. Hu et al. [271] designed a new co-estimation
the lithium-ion battery [261]. A new ultrasonic transmission based hierarchy, which can jointly estimate the SOC, SOH and SOP of
probe for the SOC has been introduced in Ref. [262]. Besides, lithium-ion batteries. Their method significantly improves the estima
acousto-ultrasonic guided waves have been used to monitor SOC and tion accuracy of SOC, voltage and capacity. In general, the joint state
SOH of Lithium-ion pouch batteries with built-in piezoelectric sensors estimation can improve the state estimation accuracy. Many scholars
[263]. In addition, high-precision contact-type displacement sensors can have begun to pay attention to this aspect of research, and believe it is a
be used to measure the swelling of the battery [264], which can be used popular research direction in the future.
to estimate SOC imbalance of batteries. Based on the mechanical rather
than electrical signal, incremental capacity analysis method can be used 4.5. Fault diagnosis and health management
to estimate the battery capacity fading [265]. In the long run, advanced
sensing technologies are expected to improve battery management. The Fault diagnosis is one of the key functions of BMS. The traditional
future development direction is to develop real-time, accurate and BMS is mainly based on the threshold method to detect the battery over-
robust sensors, combined with multi-sensor information fusion tech voltage, over-current, over temperature and conventional faults. The
nology, to obtain the deeper understanding of the internal structure and simple fault strategy cannot realize the high-level functions of fault
mechanism of the battery. tracing, early warning, etc. The traditional BMS mainly focuses on the
simple fault, which seldom involves the fault detection of thermal fault
4.2. Full life cycle modeling and aging behavior expression and sensor fault. Therefore, in the future, BMS fault diagnosis mainly has
the following development directions: first, based on artificial intelli
The current battery modeling and state estimation approaches have gence algorithm to achieve battery system fault detection, location,
made great progress. In terms of modeling, the most commonly used traceability, prediction, etc. The second is to establish a comprehensive
ECMs lack chemical significance, which can’t describe the dynamic fault diagnosis system for large-scale battery array, including battery
characteristics inside the battery. Therefore, it is recommended to thermal fault, electrical fault, sensor fault, BMS fault, etc. There are
establish physics-based models considering electrochemical processes some challenging problems in the health management of electric vehi
[266,267]. The data-driven models are obtained under certain training cles, such as modifying the battery parameters in the whole life cycle,
conditions, and there will be robustness problems in actual vehicle determine the battery safe charging and discharging constraints, and
operation. This requires model training under comprehensive aging achieve fast and efficient charging [272,273]. Accurate state estimation
conditions. The training data sets should consider the effects of battery and fault diagnosis algorithms are the basis of battery health manage
aging, hysteresis, different charge and discharge rates, temperature, etc. ment. The management strategy that considering the optimization of
Another important issue in battery modeling is the match of computing battery health, battery charging and discharging speed and battery
complexity and computing resources. Therefore, future research direc temperature protection can prevent the battery from overheating, pro
tion will also focus on battery model reduction and simplification. long the cycle life and improve the energy conversion efficiency [274].
With the development of electrochemical models and advanced state
4.3. Thermal field prediction and thermal behavior characteristic analysis estimation methods, future battery health state estimation methods will
be more applied in online applications and more integrated with battery
Simultaneously, some problems remain in terms of thermal field management strategies.
prediction and thermal characteristic analysis. On the one hand, the
inside of the cell is a closed space. In addition, the internal components 4.6. Advanced BMS architecture with 5G
are sensitive to the environment. On the other hand, battery aging is
affected by multiple stresses, and the interactions of the stresses are not The conventional BMS calibrates parameters based on laboratory
clear. Most of the current research only focus on partial stresses under data, which is difficult to meet the requirements of high accuracy and
ideal cycle conditions, and have not established a quantitative descrip real-time performance. Moreover, with the development of the vehicle
tion of the aging mechanism. The thermal field prediction and thermal to grid (V2G) technology and the advent of 5G high-speed information
characteristic analysis are crucial to thermal management. Currently, age, the bandwidth requirements of BMS have also put forward new
there are some researches focus on the development of the requirements. Therefore data-driven personalized battery management
electrochemical-thermal coupling model and thermal abuse model of scheme based on platform of big data and cloud computing will become
battery [268]. However, the research methods for the distribution of the future trend as shown in Fig. 7. It will break through the resource
current density in the electrodes inside the battery and the situation of limitation of traditional embedded hardware terminals, make full use of
the battery thermal runaway caused by side reactions at high tempera immense computing power and information storage, and upgrade the
tures are still in the simulation stage. The mechanism analysis and battery from traditional offline management to active online manage
research methods of thermal failure and thermal abuse still have ment. With the application of 5G technology in the fields of automotive
limitations. electronics and industrial control, it can be expected that the on-line
management scheme based on cloud computing will become the
4.4. Multi-states joint estimation future direction.
At present, the method of single state estimation is very rich, but in 5. Conclusions
the process of automobile operation, there is a coupling relationship
between the battery states, and the states may affect the model param In this paper, the current literature on battery modeling and state
eters. There are also many studies on joint estimation of the two states. estimation were reviewed systematically. Over 200 highly relevant
12
Fig. 7. Advanced BMS architecture with 5G.
journal and conference articles were selected and reviewed in this sur References
vey. The battery models including the physics-based electrochemical
models, the integral and fractional-order equivalent circuit models, and [1] Rahimi-Eichi H, Ojha U, Baronti F, et al. Battery management system: an
overview of its application in the smart grid and electric vehicles. IEEE Ind
the data-driven models were summarized. Battery state estimation ap Electron Mag 2013;7(2):4–16.
proaches were introduced from the perspectives of remaining capacity [2] Aaldering LJ, Leker J, Song CH. Analysis of technological knowledge stock and
and energy estimation, power capability prediction, lifespan and health prediction of its future development potential: the case of lithium-ion batteries.
J Clean Prod 2019;223:301–11.
prognoses and other important indicators relating to battery equaliza [3] Feng X, Ouyang M, Liu X, et al. Thermal runaway mechanism of lithium ion
tion and thermal management. Finally, the challenges and outlooks of battery for electric vehicles: a review. Energy Storage Mater 2018;10:246–67.
the research on battery modeling and state estimation were discussed. [4] Hu X, Cao D, Egardt B. Condition monitoring in advanced battery management
systems: moving horizon estimation using a reduced electrochemical model. IEEE
Technologies include advanced sensing, full life cycle modeling and ASME Trans Mechatron 2017;23(1):167–78.
aging behavior expression, thermal field prediction and thermal [5] Chaturvedi NA, Klein R, Christensen J, et al. Algorithms for advanced battery-
behavior characteristic analysis, multi-states joint estimation, fault management systems. IEEE Contr Syst Mag 2010;30(3):49–68.
[6] Wang Y, Chen Z, Zhang C. On-line remaining energy prediction: a case study in
diagnosis and health management are expected to be hotspot topics of
embedded battery management system. Appl Energy 2017;194:688–95.
research in the future. The data-driven battery management scheme [7] Zhang Q, Wang D, Yang B, et al. Electrochemical model of lithium-ion battery for
based on platform of big data and cloud computing is expected to be the wide frequency range applications. Electrochim Acta 2020:136094.
future trend. [8] Nejad S, Gladwin DT, Stone DA. A systematic review of lumped-parameter
equivalent circuit models for real-time estimation of lithium-ion battery states.
J Power Sources 2016;316:183–96.
Credit author statement [9] Seaman A, Dao TS, McPhee J. A survey of mathematics-based equivalent-circuit
and electrochemical battery models for hybrid and electric vehicle simulation.
J Power Sources 2014;256:410–23.
Yujie Wang: Conceptualization; Formal analysis; Investigation; Data [10] Wang Y, Yang D, Zhang X, et al. Probability based remaining capacity estimation
curation; Methodology; Supervision; Validation; Visualization; Writing - using data-driven and neural network model. J Power Sources 2016;315:
original draft; Writing - review & editing. Jiaqiang Tian, Zhendong Sun, 199–208.
[11] Wei J, Dong G, Chen Z. Remaining useful life prediction and state of health
Li Wang, Ruilong Xu and Mince Li: Conceptualization; Formal analysis; diagnosis for lithium-ion batteries using particle filter and support vector
Investigation; Writing - original draft. Zonghai Chen: Idea; Project regression. IEEE Trans Ind Electron 2017;65(7):5634–43.
administration; Resources; Writing - review & editing; Supervision; [12] Romero-Becerril A, Alvarez-Icaza L. Comparison of discretization methods
applied to the single-particle model of lithium-ion batteries. J Power Sources
Funding acquisition. 2011;196(23):10267–79.
[13] Grandjean T, Li L, Odio M, Widanage W. Global sensitivity analysis of the single
Declaration of competing interest particle lithium-ion battery model with electrolyte[C]. In: 2019 IEEE vehicle
power and propulsion conference (VPPC); 2019. p. 1–7.
[14] Doyle M, Fuller TF, Newman J. Modeling of galvanostatic charge and discharge of
The authors declare that they have no known competing financial the lithium/polymer/insertion cell. J Electrochem Soc 1993;140(6):1526–33.
interests or personal relationships that could have appeared to influence [15] Dao TS, Vyasarayani CP, McPhee J. Simplification and order reduction of lithium-
ion battery model based on porous-electrode theory. J Power Sources 2012;198:
the work reported in this paper. 329–37.
[16] Han X, Ouyang M, Lu L, et al. Simplification of physics-based electrochemical
Acknowledgment model for lithium ion battery on electric vehicle. Part II: pseudo-two-dimensional
model simplification and state of charge estimation. J Power Sources 2015;278:
814–25.
This work is supported partly by the National Natural Science [17] Xu M, Wang R, Zhao P, et al. Fast charging optimization for lithium-ion batteries
Foundation of China (Grant No. 61803359), partly by the University based on dynamic programming algorithm and electrochemical-thermal-capacity
fade coupled model. J Power Sources 2019;438:227015.
Synergy Innovation Program of Anhui Province (Grant No. GXXT-2019- [18] Song X, Lu Y, Wang F, et al. A coupled electro-chemo-mechanical model for all-
002), partly by the USTC Research Funds of the Double First-Class solid-state thin film Li-ion batteries: the effects of bending on battery
Initiative (Grant No. YD2350002002), and partly by the Fundamental performances. J Power Sources 2020;452:227803.
[19] Johnson VH. Battery performance models in ADVISOR. J Power Sources 2002;
Research Funds for the Central Universities (Grant No. 110(2):321–9.
WK2100100032).
13
[20] Liaw BY, Nagasubramanian G, Jungst RG, et al. Modeling of lithium ion cells—a [51] Zhu L, Sun Z, Dai H, et al. A novel modeling methodology of open circuit voltage
simple equivalent-circuit model approach. Solid State Ionics 2004;175(1-4): hysteresis for LiFePO4 batteries based on an adaptive discrete Preisach model.
835–9. Appl Energy 2015;155:91–109.
[21] Xia B, Sun Z, Zhang R, et al. A cubature particle filter algorithm to estimate the [52] Zheng L, Zhang L, Zhu J, et al. Co-estimation of state-of-charge, capacity and
state of the charge of lithium-ion batteries based on a second-order equivalent resistance for lithium-ion batteries based on a high-fidelity electrochemical
circuit model. Energies 2017;10(4):457. model. Appl Energy 2016;180:424–34.
[22] Andre D, Meiler M, Steiner K, et al. Characterization of high-power lithium-ion [53] Bao Y, Dong W, Wang D. Online internal resistance measurement application in
batteries by electrochemical impedance spectroscopy. I. Experimental lithium ion battery capacity and state of charge estimation. Energies 2018;11:
investigation. J Power Sources 2011;196(12):5334–41. 1073.
[23] Andre D, Meiler M, Steiner K, et al. Characterization of high-power lithium-ion [54] Huet F, Nogueira RP, Lailler P, et al. Investigation of the high-frequency
batteries by electrochemical impedance spectroscopy. II: Modelling. J Power resistance of a lead-acid battery. J Power Sources 2006;158(2):1012–8.
Sources 2011;196(12):5349–56. [55] Huang W, Qahouq JAA. An online battery impedance measurement method using
[24] Verbrugge M, Tate E. Adaptive state of charge algorithm for nickel metal hydride dc–dc power converter control. IEEE Trans Ind Electron 2014;61(11):5987–95.
batteries including hysteresis phenomena. J Power Sources 2004;126(1-2): [56] Zhang Y, Song W, Lin S, Feng Z. A novel model of the initial state of charge
236–49. estimation for LiFePO4 batteries. J Power Sources 2014;248:1028–33.
[25] Wang Y, Liu C, Pan R, et al. Modeling and state-of-charge prediction of lithium- [57] Xu L, Wang J, Chen Q. Kalman filtering state of charge estimation for battery
ion battery and ultracapacitor hybrids with a co-estimator. Energy 2017;121: management system based on a stochastic fuzzy neural network battery model.
739–50. Energy Convers Manag 2012;53(1):33–9.
[26] Zhang X, Lu J, Yuan S, et al. A novel method for identification of lithium-ion [58] Mastali M, Vazquez-Arenas J, Fraser R, et al. Battery state of the charge
battery equivalent circuit model parameters considering electrochemical estimation using Kalman filtering. J Power Sources 2013;239:294–307.
properties. J Power Sources 2017;345:21–9. [59] Yu Z, Huai R, Xiao L. State-of-charge estimation for lithium-ion batteries using a
[27] Brand J, Zhang Z, Agarwal RK. Extraction of battery parameters of the equivalent kalman filter based on local linearization. Energies 2015;8(8):7854–73.
circuit model using a multi-objective genetic algorithm. J Power Sources 2014; [60] Dong G, Wei J, Chen Z. Kalman filter for onboard state of charge estimation and
247:729–37. peak power capability analysis of lithium-ion batteries. J Power Sources 2016;
[28] Gao P, Zhang C, Wen G. Equivalent circuit model analysis on electrochemical 328:615–26.
impedance spectroscopy of lithium metal batteries. J Power Sources 2015;294: [61] Wei Z, Zhao J, Ji D, et al. A multi-timescale estimator for battery state of charge
67–74. and capacity dual estimation based on an online identified model. Appl Energy
[29] Jang J, Yoo J. Equivalent circuit evaluation method of lithium polymer battery 2017;204:1264–74.
using bode plot and numerical analysis. IEEE Trans Energy Convers 2011;26(1): [62] Xiong R, Sun F, Chen Z, He H. A data-driven multi-scale extended kalman filtering
290–8. based parameter and state estimation approach of lithium-ion olymer battery in
[30] Freeborn TJ, Maundy B, Elwakil AS. Fractional-order models of supercapacitors, electric vehicles. Appl Energy 2014;113:463–76.
batteries and fuel cells: a survey. Mater Renew Sustain Energy 2015;4(3):9. [63] Chen Z, Fu Y, Mi CC. State of charge estimation of lithium-ion batteries in electric
[31] Yang Q, Xu J, Cao B, et al. A simplified fractional order impedance model and drive vehicles using extended Kalman filtering. IEEE Trans Veh Technol 2012;62
parameter identification method for lithium-ion batteries. PloS One 2017;12(2): (3):1020–30.
e0172424. [64] Lee J, Nam O, Cho BH. Li-ion battery SOC estimation method based on the
[32] Zou Y, Li SE, Shao B, et al. State-space model with non-integer order derivatives reduced order extended Kalman filtering. J Power Sources 2007;174(1):9–15.
for lithium-ion battery. Appl Energy 2016;161:330–6. [65] Dai H, Wei X, Sun Z, et al. Online cell SOC estimation of Li-ion battery packs using
[33] Mu H, Xiong R, Zheng H, et al. A novel fractional order model based state-of- a dual time-scale Kalman filtering for EV applications. Appl Energy 2012;95:
charge estimation method for lithium-ion battery. Appl Energy 2017;207:384–93. 227–37.
[34] Xiong R, Tian J, Shen W, et al. A novel fractional order model for state of charge [66] Hu C, Youn BD, Chung J. A multiscale framework with extended Kalman filter for
estimation in lithium ion batteries. IEEE Trans Veh Technol 2018;68(5):4130–9. lithium-ion battery SOC and capacity estimation. Appl Energy 2012;92:694–704.
[35] Wang Y, Gao G, Li X, et al. A fractional-order model-based state estimation [67] Zhang C, Jiang J, Zhang W, et al. Estimation of state of charge of lithium-ion
approach for lithium-ion battery and ultra-capacitor hybrid power source system batteries used in HEV using robust extended Kalman filtering. Energies 2012;5(4):
considering load trajectory. J Power Sources 2020;449:227543. 1098–115.
[36] Hu X, Yuan H, Zou C, et al. Co-estimation of state of charge and state of health for [68] Lee KT, Dai MJ, Chuang CC. Temperature-compensated model for lithium-ion
lithium-ion batteries based on fractional-order calculus. IEEE Trans Veh Technol polymer batteries with extended Kalman filter state-of-charge estimation for an
2018;67(11):10319–29. implantable charger. IEEE Trans Ind Electron 2017;65(1):589–96.
[37] Tian J, Xiong R, Yu Q. Fractional-order model-based incremental capacity [69] Bizeray AM, Zhao S, Duncan SR, et al. Lithium-ion battery thermal-
analysis for degradation state recognition of lithium-ion batteries. IEEE Trans Ind electrochemical model-based state estimation using orthogonal collocation and a
Electron 2018;66(2):1576–84. modified extended Kalman filter. J Power Sources 2015;296:400–12.
[38] Yang R, Xiong R, He H, et al. A fractional-order model-based battery external [70] Ramadan HS, Becherif M, Claude F. Extended kalman filter for accurate state of
short circuit fault diagnosis approach for all-climate electric vehicles application. charge estimation of lithium-based batteries: a comparative analysis. Int J
J Clean Prod 2018;187:950–9. Hydrogen Energy 2017;42(48):29033–46.
[39] Wang QK, He YJ, Shen JN, et al. A unified modeling framework for lithium-ion [71] Pan H, Lü Z, Lin W, et al. State of charge estimation of lithium-ion batteries using
batteries: an artificial neural network based thermal coupled equivalent circuit a grey extended Kalman filter and a novel open-circuit voltage model. Energy
model approach. Energy 2017;138:118–32. 2017;138:764–75.
[40] Zhang C, Zhu Y, Dong G, et al. Data-driven lithium-ion battery states estimation [72] Zou C, Manzie C, Nesic D, et al. Multi-time-scale observer design for state-of-
using neural networks and particle filtering. Int J Energy Res 2019;43(14): charge and state-of-health of a lithium-ion battery. J Power Sources 2016:
8230–41. 121–30.
[41] Dong G, Zhang X, Zhang C, et al. A method for state of energy estimation of [73] He H, Xiong R, Zhang X, et al. State-of-charge estimation of the lithium-ion
lithium-ion batteries based on neural network model. Energy 2015;90:879–88. battery using an adaptive extended Kalman filter based on an improved Thevenin
[42] Li S, He H, Li J. Big data driven lithium-ion battery modeling method based on model. IEEE Trans Veh Technol 2011;60(4):1461–9.
SDAE-ELM algorithm and data pre-processing technology. Appl Energy 2019;242: [74] He H, Xiong R, Guo H. Online estimation of model parameters and state-of-charge
1259–73. of LiFePO4 batteries in electric vehicles. Appl Energy 2012;89(1):413–20.
[43] Vigneshwaran T, Oscar XG, Matias Q, et al. A three dimensional kinetic Monte [75] Xiong R, He H, Sun F, et al. Evaluation on state of charge estimation of batteries
Carlo model for simulating the carbon/sulfur mesostructural evolutions of with adaptive extended Kalman filter by experiment approach. IEEE Trans Veh
discharging lithium sulfur batteries. Energy Storage Mater 2020;24:472–85. Technol 2012;62(1):108–17.
[44] Plett GL. Battery management systems, Volume II: equivalent-circuit methods [76] Xiong R, Gong X, Mi CC, et al. A robust state-of-charge estimator for multiple
[M]. Artech House; 2015. types of lithium-ion batteries using adaptive extended Kalman filter. J Power
[45] How DNT, Hannan MA, Lipu MSH, et al. State of charge estimation for lithium- Sources 2013;243:805–16.
ion batteries using model-based and data-driven methods: a review. IEEE Access [77] Sepasi S, Ghorbani R, Liaw BY. A novel on-board state-of-charge estimation
2019;7:136116–36. method for aged Li-ion batteries based on model adaptive extended Kalman filter.
[46] Pop V, Bergveld H, het Veld J O, et al. Modeling battery behavior for accurate J Power Sources 2014;245:337–44.
state-of-charge indication. J Electrochem Soc 2006;153:A2013–22. [78] Zhu Q, Xu M, Liu W, et al. A state of charge estimation method for lithium-ion
[47] Ali MU, Zafar A, Nengroo SH, et al. Towards a smarter battery management batteries based on fractional order adaptive extended kalman filter. Energy 2019;
system for electric vehicle applications: a critical review of lithium-ion battery 187:115880.
state of charge estimation. Energies 2019;12(3):446. [79] Jiang B, Dai H, Wei X, et al. Joint estimation of lithium-ion battery state of charge
[48] Klintberg A, Zou C, Fridholm B, et al. Kalman filter for adaptive learning of two- and capacity within an adaptive variable multi-timescale framework considering
dimensional look-up tables applied to OCV-curves for aged battery cells. Contr current measurement offset. Appl Energy 2019:113619.
Eng Pract 2019;84:230–7. [80] Seo BH, Nguyen TH, Lee DC, et al. Condition monitoring of lithium polymer
[49] Xing Y, He W, Pecht M, et al. State of charge estimation of lithium-ion batteries batteries based on a sigma-point Kalman filter. J Power Electron 2012;12(5):
using the open-circuit voltage at various ambient temperatures. Appl Energy 778–86.
2014;113:106–15. [81] Bi Y, Choe SY. An adaptive sigma-point Kalman filter with state equality
[50] Dong G, Wei J, Zhang C, et al. Online state of charge estimation and open circuit constraints for online state-of-charge estimation of a Li (NiMnCo) O2/Carbon
voltage hysteresis modeling of LiFePO4 battery using invariant imbedding battery using a reduced-order electrochemical model. Appl Energy 2020;258:
method. Appl Energy 2016;162:163–71. 113925.
14
[82] Zhang J, Xia C. State-of-charge estimation of valve regulated lead acid battery [112] Wang D, Yang F, Tsui KL, et al. Remaining useful life prediction of lithium-ion
based on multi-state Unscented Kalman Filter. Int J Electr Power Energy Syst batteries based on spherical cubature particle filter. IEEE Trans Instrum Meas
2011;33(3):472–6. 2016;65(6):1282–91.
[83] Chen Z, Yang L, Zhao X, et al. Online state of charge estimation of Li-ion battery [113] Guo R, Gan Q, Zhang J, et al. Huber cubature particle filter and online state
based on an improved unscented Kalman filter approach. Appl Math Model 2019; estimation. Proc IME J Syst Contr Eng 2017;231(3):158–67.
70:532–44. [114] Xia B, Sun Z, Zhang R, et al. A comparative study of three improved algorithms
[84] Aung H, Low KS. Temperature dependent state-of-charge estimation of lithium based on particle filter algorithms in soc estimation of lithium ion batteries.
ion battery using dual spherical unscented Kalman filter. IET Power Electron Energies 2017;10(8):1149.
2015;8(10):2026–33. [115] Zhang K, Ma J, Zhao X, et al. State of charge estimation for lithium battery based
[85] Gholizade-Narm H, Charkhgard M. Lithium-ion battery state of charge estimation on adaptively weighting cubature particle filter. IEEE Access 2019;7:166657–66.
based on square-root unscented Kalman filter. IET Power Electron 2013;6(9): [116] Luenberger D. An introduction to observers. IEEE Trans Automat Contr 1971;16
1833–41. (6):596–602.
[86] Aung H, Low KS, Goh ST. State-of-charge estimation of lithium-ion battery using [117] Luenberger D. Observers for multivariable systems. IEEE Trans Automat Contr
square root spherical unscented Kalman filter (Sqrt-UKFST) in nanosatellite. IEEE 1966;11(2):190–7.
Trans Power Electron 2014;30(9):4774–83. [118] Hu X, Sun F, Zou Y. Estimation of state of charge of a lithium-ion battery pack for
[87] Zeng M, Zhang P, Yang Y, et al. SOC and SOH joint estimation of the power electric vehicles using an adaptive Luenberger observer. Energies 2010;3(9):
batteries based on fuzzy unscented kalman filtering algorithm. Energies 2019;12 1586–603.
(16):3122. [119] Wang B, Liu Z, Li SE, et al. State-of-charge estimation for lithium-ion batteries
[88] Sun F, Hu X, Zou Y, et al. Adaptive unscented Kalman filtering for state of charge based on a nonlinear fractional model. IEEE Trans Contr Syst Technol 2016;25(1):
estimation of a lithium-ion battery for electric vehicles. Energy 2011;36(5): 3–11.
3531–40. [120] Kim D, Koo K, Jeong JJ, et al. Second-order discrete-time sliding mode observer
[89] Liu G, Xu C, Li H, et al. State of charge and online model parameters co-estimation for state of charge determination based on a dynamic resistance li-ion battery
for liquid metal batteries. Appl Energy 2019;250:677–84. model. Energies 2013;6(10):5538–51.
[90] Partovibakhsh M, Liu G. An adaptive unscented Kalman filtering approach for [121] Chen X, Shen W, Cao Z, et al. Adaptive gain sliding mode observer for state of
online estimation of model parameters and state-of-charge of lithium-ion charge estimation based on combined battery equivalent circuit model. Comput
batteries for autonomous mobile robots. IEEE Trans Contr Syst Technol 2014;23 Chem Eng 2014;64:114–23.
(1):357–63. [122] Chen X, Shen W, Cao Z, et al. A novel approach for state of charge estimation
[91] Peng S, Chen C, Shi H, et al. State of charge estimation of battery energy storage based on adaptive switching gain sliding mode observer in electric vehicles[J].
systems based on adaptive unscented Kalman filter with a noise statistics J Power Sources 2014;246:667–78.
estimator. Ieee Access 2017;5:13202–12. [123] Kim D, Goh T, Park M, et al. Fuzzy sliding mode observer with grey prediction for
[92] Lim J. CDKF approach for estimating a static parameter of carrier frequency offset the estimation of the state-of-charge of a lithium-ion battery. Energies 2015;8
based on nonlinear measurement equations in OFDM systems. Nonlinear Dynam (11):12409–28.
2014;78(1):703–11. [124] Chen X, Shen W, Dai M, et al. Robust adaptive sliding-mode observer using RBF
[93] Sangwan V, Kumar R, Rathore AK. State-of-charge estimation for li-ion battery neural network for lithium-ion battery state of charge estimation in electric
using extended Kalman filter (EKF) and central difference Kalman filter (CDKF) vehicles. IEEE Trans Veh Technol 2015;65(4):1936–47.
[C]. In: 2017 IEEE industry applications society annual meeting. IEEE; 2017. [125] Zhong Q, Zhong F, Cheng J, et al. State of charge estimation of lithium-ion
p. 1–6. batteries using fractional order sliding mode observer. ISA Trans 2017;66:
[94] Xuan DJ, Shi Z, Chen J, et al. Real-time estimation of state-of-charge in lithium- 448–59.
ion batteries using improved central difference transform method. J Clean Prod [126] Huangfu Y, Xu J, Zhao D, et al. A novel battery state of charge estimation method
2020;252:119787. based on a super-twisting sliding mode observer. Energies 2018;11(5):1211.
[95] Arasaratnam I, Haykin S. Cubature kalman filters. IEEE Trans Automat Contr [127] Xu J, Mi CC, Cao B, et al. The state of charge estimation of lithium-ion batteries
2009;54(6):1254–69. based on a proportional-integral observer. IEEE Trans Veh Technol 2013;63(4):
[96] Peng J, Luo J, He H, et al. An improved state of charge estimation method based 1614–21.
on cubature Kalman filter for lithium-ion batteries. Appl Energy 2019;253: [128] Tang X, Wang Y, Chen Z. A method for state-of-charge estimation of LiFePO4
113520. batteries based on a dual-circuit state observer. J Power Sources 2015;296:23–9.
[97] Linghu J, Kang L, Liu M, et al. Estimation for state-of-charge of lithium-ion [129] Xie J, Ma J, Sun Y, et al. Estimating the state-of-charge of lithium-ion batteries
battery based on an adaptive high-degree cubature Kalman filter. Energy 2019; using an H-infinity observer with consideration of the hysteresis characteristic.
189:116204. J Power Electron 2016;16(2):643–53.
[98] Liu Z, Dang X, Jing B, et al. A novel model-based state of charge estimation for [130] Zhu Q, Xiong N, Yang ML, et al. State of charge estimation for lithium-ion battery
lithium-ion battery using adaptive robust iterative cubature Kalman filter. Elec based on nonlinear observer: an h∞ method. Energies 2017;10(5):679.
Power Syst Res 2019;177:105951. [131] Zhu Q, Li L, Hu X, et al. H∞-Based nonlinear observer design for state of charge
[99] Cui X, Jing Z, Luo M, et al. A new method for state of charge estimation of estimation of lithium-ion battery with polynomial parameters. IEEE Trans Veh
lithium-ion batteries using square root cubature Kalman filter. Energies 2018;11 Technol 2017;66(12):10853–65.
(1):209. [132] Liu C, Zhu Q, Li L, et al. A state of charge estimation method based on H∞
[100] Doucet A, De Freitas N, Gordon N. An introduction to sequential Monte Carlo observer for switched systems of lithium-ion nickel–manganese–cobalt batteries.
methods[M]. In: Sequential Monte Carlo methods in practice. New York, NY: IEEE Trans Ind Electron 2017;64(10):8128–37.
Springer; 2001. p. 3–14. [133] Wei Z, Zou C, Leng F, et al. Online model identification and state-of-charge
[101] Gordon NJ, Salmond DJ, Smith AFM. Novel approach to nonlinear/non-Gaussian estimate for lithium-ion battery with a recursive total least squares-based
Bayesian state estimation[C]. In: IEE proceedings F (radar and signal processing). observer. IEEE Trans Ind Electron 2017;65(2):1336–46.
IET Digital Library, vol. 140; 1993. p. 107–13. 2. [134] Wei Z, Meng S, Xiong B, et al. Enhanced online model identification and state of
[102] Nummiaro K, Koller-Meier E, Van Gool L. An adaptive color-based particle filter. charge estimation for lithium-ion battery with a FBCRLS based observer. Appl
Image Vis Comput 2003;21(1):99–110. Energy 2016;181:332–41.
[103] Tulsyan A, Tsai Y, Gopaluni RB, et al. State-of-charge estimation in lithium-ion [135] Ma Y, Li B, Li G, et al. A nonlinear observer approach of SOC estimation based on
batteries: a particle filter approach. J Power Sources 2016;331:208–23. hysteresis model for lithium-ion battery. IEEE/CAA J Autom Sin 2017;4(2):
[104] Chen Z, Sun H, Dong G, et al. Particle filter-based state-of-charge estimation and 195–204.
remaining-dischargeable-time prediction method for lithium-ion batteries. [136] Dey S, Ayalew B, Pisu P. Nonlinear robust observers for state-of-charge estimation
J Power Sources 2019;414:158–66. of lithium-ion cells based on a reduced electrochemical model. IEEE Trans Contr
[105] Schwunk S, Armbruster N, Straub S, et al. Particle filter for state of charge and Syst Technol 2015;23(5):1935–42.
state of health estimation for lithium–iron phosphate batteries. J Power Sources [137] Li W, Liang L, Liu W, et al. State of charge estimation of lithium-ion batteries
2013;239:705–10. using a discrete-time nonlinear observer. IEEE Trans Ind Electron 2017;64(11):
[106] Wang Y, Zhang C, Chen Z. A method for state-of-charge estimation of LiFePO4 8557–65.
batteries at dynamic currents and temperatures using particle filter. J Power [138] Xia B, Chen C, Tian Y, et al. State of charge estimation of lithium-ion batteries
Sources 2015;279:306–11. based on an improved parameter identification method. Energy 2015;90:
[107] Liu X, Chen Z, Zhang C, et al. A novel temperature-compensated model for power 1426–34.
Li-ion batteries with dual-particle-filter state of charge estimation. Appl Energy [139] Dang X, Yan L, Xu K, et al. Open-circuit voltage-based state of charge estimation
2014;123:263–72. of lithium-ion battery using dual neural network fusion battery model.
[108] Van Der Merwe R, Doucet A, De Freitas N, et al. The unscented particle filter[C]. Electrochim Acta 2016;188:356–66.
Adv Neural Inf Process Syst 2001:584–90. [140] Sun F, Xiong R, He H. A systematic state-of-charge estimation framework for
[109] He Y, Liu XT, Zhang CB, et al. A new model for State-of-Charge (SOC) estimation multi-cell battery pack in electric vehicles using bias correction technique. Appl
for high-power Li-ion batteries. Appl Energy 2013;101:808–14. Energy 2016;162:1399–409.
[110] Wang Y, Chen Z. A framework for state-of-charge and remaining discharge time [141] Tong S, Lacap JH, Park JW. Battery state of charge estimation using a load-
prediction using unscented particle filter. Appl Energy 2020;260:114324. classifying neural network. J Energy Storage 2016;7:236–43.
[111] Shen Y. Hybrid unscented particle filter based state-of-charge determination for [142] Chemali E, Kollmeyer PJ, Preindl M, et al. Long short-term memory networks for
lead-acid batteries. Energy 2014;74:795–803. accurate state-of-charge estimation of Li-ion batteries. IEEE Trans Ind Electron
2017;65(8):6730–9.
15
[143] Hossain Lipu MS, Hannan MA, Hussain A, et al. Optimal BP neural network [175] Pan R, Wang Y, Zhang X, et al. Power capability prediction for lithium-ion
algorithm for state of charge estimation of lithium-ion battery using PSO with batteries based on multiple constraints analysis [J]. Electrochim Acta 2017;238:
PCA feature selection. J Renew Sustain Energy 2017;9(6):064102. 120–33.
[144] Chemali E, Kollmeyer PJ, Preindl M, et al. State-of-charge estimation of Li-ion [176] Wei Z, Meng S, Tseng KJ, et al. An adaptive model for vanadium redox flow
batteries using deep neural networks: a machine learning approach. J Power battery and its application for online peak power estimation. J Power Sources
Sources 2018;400:242–55. 2017;344:195–207.
[145] Chen J, Ouyang Q, Xu C, et al. Neural network-based state of charge observer [177] Zhang W, Shi W, Ma Z. Adaptive unscented Kalman filter based state of energy
design for lithium-ion batteries. IEEE Trans Contr Syst Technol 2017;26(1): and power capability estimation approach for lithium-ion battery. J Power
313–20. Sources 2015;289:50–62.
[146] Lipu MSH, Hannan MA, Hussain A, et al. State of charge estimation for lithium- [178] Sun F, Xiong R, He H. Estimation of state-of-charge and state-of-power capability
ion battery using recurrent NARX neural network model based lighting search of lithium-ion battery considering varying health conditions. J Power Sources
algorithm. IEEE Access 2018;6:28150–61. 2014;259:166–76.
[147] Xia B, Cui D, Sun Z, et al. State of charge estimation of lithium-ion batteries using [179] Wang Y, Zhang X, Liu C, et al. Multi-timescale power and energy assessment of
optimized Levenberg-Marquardt wavelet neural network. Energy 2018;153: lithium-ion battery and supercapacitor hybrid system using extended Kalman
694–705. filter. J Power Sources 2018;389:93–105.
[148] Yang F, Li W, Li C, et al. State-of-charge estimation of lithium-ion batteries based [180] Tang X, Wang Y, Yao K, et al. Model migration based battery power capability
on gated recurrent neural network. Energy 2019;175:66–75. evaluation considering uncertainties of temperature and aging. J Power Sources
[149] Yang F, Song X, Xu F, et al. State-of-charge estimation of lithium-ion batteries via 2019;440:227141.
long short-term memory network. IEEE Access 2019;7:53792–9. [181] Zou C, Klintberg A, Wei Z, Fridholm B, Wik T, Egardt B. Power capability
[150] Souradip M. Fuzzy logic-based learning system and estimation of state-of-charge prediction for lithium-ion batteries using economic nonlinear model predictive
of lead-acid battery. Eng Appl Artif Intell 2006;19:479–85. control. J Power Sources 2018;396:580–9.
[151] Burgos C, Saez D, Orchard ME, et al. Fuzzy modelling for the state-of-charge [182] Fleischer C, Waag W, Bai Z, et al. Adaptive on-line state-of-available-power
estimation of lead-acid batteries. J Power Sources 2015;274:355–66. prediction of lithium-ion batteries. J Power Electron 2013;13(4):516–27.
[152] Lee DT, Shiah SJ, Lee CM, et al. State-of-charge estimation for electric scooters by [183] Hu X, Xu L, Lin X, et al. Battery lifetime prognostics. Joule 2020;4:310–46.
using learning mechanisms. IEEE Trans Veh Technol 2007;56(2):544–56. [184] Ng KS, Moo C, Chen Y, et al. Enhanced coulomb counting method for estimating
[153] Li IH, Wang WY, Su SF, et al. A merged fuzzy neural network and its applications state-of-charge and state-of-health of lithium-ion batteries. Appl Energy 2009;86
in battery state-of-charge estimation. IEEE Trans Energy Convers 2007;22(3): (9):1506–11.
697–708. [185] Kim T, Qiao W, Qu L. Online SOC and SOH estimation for multi cell lithium-ion
[154] Ting TO, Man KL, Lim EG, et al. Tuning of Kalman filter parameters via genetic batteries based on an adaptive hybrid battery model and sliding-mode observer
algorithm for state-of-charge estimation in battery management system. Sci World [C]. In: 2013 IEEE energy conversion congress and exposition. IEEE; 2013.
J 2014:176052. p. 292–8.
[155] Chen L, Wang Z, Lü Z, et al. A novel state-of-charge estimation method of lithium- [186] Blanke H, Bohlen O, Buller S, et al. Impedance measurements on lead-acid
ion batteries combining the grey model and genetic algorithms. IEEE Trans Power batteries for state-of-charge, state-of-health and cranking capability prognosis in
Electron 2017;33(10):8797–807. electric and hybrid electric vehicles. J Power Sources 2005;144(2):418–25.
[156] Anton J, Nieto P, Viejo C, Vilan J. Support vector machines used to estimate the [187] Galeotti M, Giammanco C, Cina L, et al. Diagnostic methods for the evaluation of
battery state of charge. IEEE Trans Power Electron 2013;28:5919–26. the state of health (SOH) of NiMH batteries through electrochemical impedance
[157] Hu JN, Hu JJ, Lin HB, et al. State-of-charge estimation for battery management spectroscopy[C]. In: 2014 IEEE 23rd international symposium on industrial
system using optimized support vector machine for regression. J Power Sources electronics (ISIE). IEEE; 2014. p. 1641–6.
2014;269:682–93. [188] Dubarry M, Svoboda V, Hwu R, et al. Capacity and power fading mechanism
[158] Meng J, Luo G, Gao F. Lithium polymer battery state-of-charge estimation based identification from a commercial cell evaluation. J Power Sources 2007;165(2):
on adaptive unscented Kalman filter and support vector machine. IEEE Trans 566–72.
Power Electron 2015;31(3):2226–38. [189] Weng C, Cui Y, Sun J, et al. On-board state of health monitoring of lithium-ion
[159] Deng Z, Yang L, Cai Y, et al. Online available capacity prediction and state of batteries using incremental capacity analysis with support vector regression.
charge estimation based on advanced data-driven algorithms for lithium iron J Power Sources 2013;235(235):36–44.
phosphate battery. Energy 2016;112:469–80. [190] Merla Y, Wu B, Yufit V, et al. Novel application of differential thermal
[160] Mamadou K, Lemaire E, Delaille A, et al. Definition of a state-of-energy indicator voltammetry as an in-depth state-of-health diagnosis method for lithium-ion
(SoE) for electrochemical storage devices: application for energetic availability batteries. J Power Sources 2016:308–19.
forecasting. J Electrochem Soc 2012;159(8):A1298–307. [191] Wu B, Yufit V, Merla Y, et al. Differential thermal voltammetry for tracking of
[161] Liu X, Wu J, Zhang C, et al. A method for state of energy estimation of lithium-ion degradation in lithium-ion batteries. J Power Sources 2015:495–501.
batteries at dynamic currents and temperatures. J Power Sources 2014;270: [192] Maher K, Yazami R. A study of lithium ion batteries cycle aging by
151–7. thermodynamics techniques. J Power Sources 2014:527–33.
[162] Wang Y, Zhang C, Chen Z. A method for joint estimation of state-of-charge and [193] Berecibar M, Garmendia M, Gandiaga I, et al. State of health estimation algorithm
available energy of LiFePO4 batteries. Appl Energy 2014;135:81–7. of LiFePO4 battery packs based on differential voltage curves for battery
[163] He HW, Zhang YZ, Xiong R, et al. A novel Gaussian model based battery state management system application. Energy 2016:784–96.
estimation approach: state-of-Energy. Appl Energy 2015;151:41–8. [194] Wang L, Pan C, Liu L, et al. On-board state of health estimation of LiFePO4 battery
[164] Zhang W, Shi W, Ma Z. Adaptive unscented Kalman filter based state of energy pack through differential voltage analysis. Appl Energy 2016:465–72.
and power capability estimation approach for lithium-ion battery. J Power [195] Sommer L, Raghavan A, Kiesel P, et al. Monitoring of intercalation stages in
Sources 2015;289:50–62. lithium-ion cells over charge-discharge cycles with fiber optic sensors.
[165] Zhang Y, Xiong R, He H, et al. Lithium-ion battery pack state of charge and state J Electrochem Soc 2015;162(14).
of energy estimation algorithms using a hardware-in-the-loop validation. IEEE [196] Peabody C, Arnold CB. The role of mechanically induced separator creep in
Trans Power Electron 2016;32(6):4421–31. lithium-ion battery capacity fade[J]. J Power Sources 2011;196(19):8147–53.
[166] Zheng L, Zhu J, Wang G, et al. Novel methods for estimating lithium-ion battery [197] Cannarella J, Arnold CB. State of health and charge measurements in lithium-ion
state of energy and maximum available energy. Appl Energy 2016;178:1–8. batteries using mechanical stress. J Power Sources 2014:7–14.
[167] Dong G, Chen Z, Wei J, et al. An online model-based method for state of energy [198] Deisenroth MP, Turner R, Huber MF, et al. Robust filtering and smoothing with
estimation of lithium-ion batteries using dual filters. J Power Sources 2016;301: Gaussian processes. IEEE Trans Automat Contr 2012;57(7):1865–71.
277–86. [199] Doucet A. On sequential Monte Carlo methods for Bayesian filtering. Stat Comput
[168] Lin C, Mu H, Xiong R, et al. Multi-model probabilities based state fusion 1998;10(3):197–208.
estimation method of lithium-ion battery for electric vehicles: state-of-energy. [200] Burgess WL. Valve Regulated Lead Acid battery float service life estimation using
Appl Energy 2017;194:560–8. a Kalman filter. J Power Sources 2009;191(1):16–21.
[169] Wang Y, Pan R, Liu C, et al. Power capability evaluation for lithium iron [201] Wassiliadis N, Adermann J, Frericks A, et al. Revisiting the dual extended Kalman
phosphate batteries based on multi-parameter constraints estimation. J Power filter for battery state-of-charge and state-of-health estimation: a use-case life
Sources 2018;374:12–23. cycle analysis. J Energy Storage 2018:73–87.
[170] Hunt G. USABC electric vehicle battery test procedures manual[J]. Washington, [202] Xue Z, Zhang Y, Cheng C, et al. Remaining useful life prediction of lithium-ion
DC, USA: United States Department of Energy; 1996. batteries with adaptive unscented kalman filter and optimized support vector
[171] Plett GL. High-performance battery-pack power estimation using a dynamic cell regression. Neurocomputing 2019:95–102.
model. IEEE Trans Veh Technol 2004;53(5):1586–93. [203] Andre D, Nuhic A, Soczkaguth T, et al. Comparative study of a structured neural
[172] Feng T, Yang L, Zhao X, et al. Online identification of lithium-ion battery network and an extended Kalman filter for state of health determination of
parameters based on an improved equivalent-circuit model and its lithium-ion batteries in hybrid electricvehicles. Eng Appl Artif Intell 2013;26(3):
implementation on battery state-of-power prediction. J Power Sources 2015;281: 951–61.
192–203. [204] Goebel K, Saha B, Saxena A, et al. Prognostics in battery health management.
[173] Dong G, Wei J, Chen Z. Kalman filter for onboard state of charge estimation and IEEE Instrum Meas Mag 2008;11(4):33–40.
peak power capability analysis of lithium-ion batteries. J Power Sources 2016; [205] Arulampalam MS, Maskell S, Gordon NJ, et al. A tutorial on particle filters for
328:615–26. online nonlinear/non-Gaussian Bayesian tracking. IEEE Trans Signal Process
[174] Xiong R, He H, Sun F, et al. Online estimation of peak power capability of Li-ion 2002;50(2):174–88.
batteries in electric vehicles by a hardware-in-loop approach. Energies 2012;5(5):
1455–69.
16
[206] Tang X, Liu K, Wang X, et al. Real-time aging trajectory prediction using a base [238] Wang P, Zhang X, Yang L, et al. Real-time monitoring of internal temperature
model-oriented gradient-correction particle filter for Lithium-ion batteries. evolution of the lithium-ion coin cell battery during the charge and discharge
J Power Sources 2019;440:227118. process. Extreme Mech Lett 2016;9:459–66.
[207] Doucet A, Godsill SJ, Andrieu C, et al. On sequential Monte Carlo sampling [239] Raijmakers LHJ, Danilov DL, Eichel RA, et al. A review on various temperature-
methods for Bayesian filtering. Stat Comput 2000;10(3):197–208. indication methods for Li-ion batteries. Appl Energy 2019;240:918–45.
[208] Miao Q, Xie L, Cui H, et al. Remaining useful life prediction of lithium-ion battery [240] Xie Y, He X, Hu X, et al. An improved resistance-based thermal model for a pouch
with unscented particle filter technique. Microelectron Reliab 2013;53(6): lithium-ion battery considering heat generation of posts. Appl Therm Eng 2020;
805–10. 164:114455.
[209] Hu C, Jain G, Tamirisa PA, et al. Method for estimating capacity and predicting [241] Lin X, Perez HE, Siegel JB, et al. Online parameterization of lumped thermal
remaining useful life of lithium-ion battery. Appl Energy 2014:182–9. dynamics in cylindrical lithium ion batteries for core temperature estimation and
[210] Zhang L, Mu Z, Sun C, et al. Remaining useful life prediction for lithium-ion health monitoring. IEEE Trans Contr Syst Technol 2012;21(5):1745–55.
batteries based on exponential model and particle filter. IEEE Access 2018: [242] Wang M, Li HX. Real-time estimation of temperature distribution for cylindrical
17729–40. lithium-ion batteries under boundary cooling. IEEE Trans Ind Electron 2016;64
[211] Wang D, Yang F, Tsui K, et al. Remaining useful life prediction of lithium-ion (3):2316–24.
batteries based on spherical cubature particle filter. IEEE Trans Instrum Meas [243] Kim Y, Mohan S, Siegel JB, et al. The estimation of temperature distribution in
2016;65(6):1282–91. cylindrical battery cells under unknown cooling conditions. IEEE Trans Contr Syst
[212] Wu J, Zhang C, Chen Z. An online method for lithium-ion battery remaining Technol 2014;22(6):2277–86.
useful life estimation using importance sampling and neural networks. Appl [244] Dai H, Zhu L, Zhu J, et al. Adaptive Kalman filtering based internal temperature
Energy 2016;173:134–40. estimation with an equivalent electrical network thermal model for hard-cased
[213] Zhang H, Miao Q, Zhang X, et al. An improved unscented particle filter approach batteries. J Power Sources 2015;293:351–65.
for lithium-ion battery remaining useful life prediction. Microelectron Reliab [245] Parhizi M, Ahmed MB, Jain A. Determination of the core temperature of a Li-ion
2018;81(81):288–98. cell during thermal runaway. J Power Sources 2017;370:27–35.
[214] Thomas EV, Bloom I, Christophersen JP, et al. Modeling memoryless degradation [246] Schmidt JP, Arnold S, Loges A, et al. Measurement of the internal cell temperature
under variable stress. J Qual Technol 2019;51(3):284–99. via impedance: evaluation and application of a new method. J Power Sources
[215] Wang JS, Liu P, Hicksgarner J, et al. Cycle-life model for graphite-LiFePO4 cells 2013;243:110–7.
[J]. J Power Sources 2011;196(8):3942–8. [247] Srinivasan R, Carkhuff BG, Butler MH, et al. Instantaneous measurement of the
[216] Schmalstieg J, Kabitz S, Ecker M, et al. A holistic aging model for Li(NiMnCo)O2 internal temperature in lithium-ion rechargeable cells. Electrochim Acta 2011;56
based 18650 lithium-ion batteries. J Power Sources 2014:325–34. (17):6198–204.
[217] Smiley A, Harrison WK, Plett GL, et al. Postprocessing the outputs of an [248] Beelen H, Raijmakers LHJ, Donkers MCF, et al. A comparison and accuracy
interacting multiple-model Kalman filter using a Markovian trellis to estimate analysis of impedance-based temperature estimation methods for Li-ion batteries.
parameter values of aged Li-ion cells. J Energy Storage 2020;27:101043. Appl Energy 2016;175:128–40.
[218] Zhang Y, Xiong R, He H, et al. State of charge-dependent aging mechanisms in [249] Schindler S, Bauer M, Petzl M, et al. Voltage relaxation and impedance
graphite/Li(NiCoAl)O2 cells: capacity loss modeling and remaining useful life spectroscopy as in-operando methods for the detection of lithium plating on
prediction. Appl Energy 2019;255(1):113818. graphitic anodes in commercial lithium-ion cells. J Power Sources 2016;304:
[219] Tang X, Liu K, Wang X, et al. Model migration neural network for predicting 170–80.
battery aging trajectories. IEEE Trans Transport Electrification 2020:363–74. [250] Barai A, Chouchelamane GH, Guo Y, et al. A study on the impact of lithium-ion
[220] Chang Y, Fang H, Zhang Y, et al. A new hybrid method for the prediction of the cell relaxation on electrochemical impedance spectroscopy. J Power Sources
remaining useful life of a lithium-ion battery. Appl Energy 2017:1564–78. 2015;280:74–80.
[221] Nuhic A, Terzimehic T, Soczkaguth T, et al. Health diagnosis and remaining useful [251] Wang X, Wei X, Chen Q, et al. Lithium-ion battery temperature on-line estimation
life prognostics of lithium-ion batteries using data-driven methods. J Power based on fast impedance calculation. J Energy Storage 2019;26:100952.
Sources 2013:680–8. [252] Beelen H, Mundaragi Shivakumar K, Raijmakers L, et al. Towards impedance-
[222] Dai H, Zhao G, Lin M, et al. A novel estimation method for the state of health of based temperature estimation for Li-ion battery packs. Int J Energy Res 2020;44
lithium-ion battery using prior knowledge-based neural network and Markov (4):2889–908.
chain. IEEE Trans Ind Electron 2019;66(10):7706–16. [253] Lin X, Perez HE, Siegel JB, et al. Robust estimation of battery system temperature
[223] Feng X, Weng C, He X, et al. Online state-of-health estimation for Li-ion battery distribution under sparse sensing and uncertainty. IEEE Trans Contr Syst Technol
using partial charging segment based on support vector machine. IEEE Trans Veh 2019:1–13.
Technol 2019;68(9):8583–92. [254] Richardson RR, Howey DA. Sensorless battery internal temperature estimation
[224] Liu K, Li Y, Hu X, et al. Gaussian process regression with automatic relevance using a Kalman filter with impedance measurement. IEEE Trans Sustain Energy
determination kernel for calendar aging prediction of lithium-ion batteries. IEEE 2015;6(4):1190–9.
Trans Industrial Inf 2020;16(6):3767–77. [255] Liu K, Li K, Peng Q, et al. Data-driven hybrid internal temperature estimation
[225] Liu K, Hu X, Wei Z, et al. Modified Gaussian process regression models for cyclic approach for battery thermal management. Complexity 2018:96428.
capacity prediction of lithium-ion batteries. IEEE Trans Transport Electrification [256] Wang S, Shang L, Li Z, et al. Online dynamic equalization adjustment of high-
2019;5(4):1225–36. power lithium-ion battery packs based on the state of balance estimation. Appl
[226] Liu K, Shang Y, Ouyang Q, et al. A data-driven approach with uncertainty Energy 2016;166:44–58.
quantification for predicting future capacities and remaining useful life of [257] Wang S, Fernandez C, Zou C, et al. A comprehensive working state monitoring
lithium-ion battery[J]. IEEE Trans Ind Electron 2020. 1-1. method for power battery packs considering state of balance and aging correction.
[227] Richardson RR, Osborne MA, Howey DA, et al. Gaussian process regression for Energy 2019;171:444–55.
forecasting battery state of health. J Power Sources 2017:209–19. [258] Bea C, Manandhar A, Kiesel P, et al. Monitoring the strain evolution of lithium-ion
[228] Pola DA, Navarrete HF, Orchard ME, et al. Particle-filtering-based discharge time battery electrodes using an optical fiber Bragg grating sensor. Energy Technol
prognosis for lithium-ion batteries with a statistical characterization of use 2016;4(7):851–5.
profiles. IEEE Trans Reliab 2015;64(2):710–20. [259] Raghavan A, Kiesel P, Sommer LW, et al. Embedded fiber-optic sensing for
[229] Dong G, Wei J, Chen Z, et al. Remaining dischargeable time prediction for accurate internal monitoring of cell state in advanced battery management
lithium-ion batteries using unscented Kalman filter. J Power Sources 2017;364: systems part 1: cell embedding method and performance. J Power Sources 2017;
316–27. 341:466–73.
[230] Qui~ nones F, Milocco RH, Real SG. Remaining discharge-time prediction for [260] Ganguli A, Saha B, Raghavan A, et al. Embedded fiber-optic sensing for accurate
batteries using the Lambert function. J Power Sources 2018;400:256–63. internal monitoring of cell state in advanced battery management systems part 2:
[231] Jinsong Y, Shuang L, Diyin T, et al. Remaining discharge time prognostics of internal cell signals and utility for state estimation. J Power Sources 2017;341:
lithium-ion batteries using Dirichlet process mixture model and particle filtering 474–82.
method. IEEE Trans Instrum Meas 2017;66(9):2317–28. [261] Jung H, Lin CF, Gerasopoulos K, et al. A buckled membrane sensor for in situ
[232] Zhang C, Yu X, Dong G, et al. A method for remaining discharge time prediction of mechanical and microstructure analysis of li-ion battery electrodes[C]. In: 2015
lithium-ion batteries under dynamic uncertainty. Int J Energy Res 2019;43(5): transducers-2015 18th international conference on solid-state sensors, actuators
1760–74. and microsystems. IEEE; 2015. p. 1953–6.
[233] Wang Q, Ping P, Zhao X, et al. Thermal runaway caused fire and explosion of [262] Gold L, Bach T, Virsik W, et al. Probing lithium-ion batteries’ state-of-charge
lithium ion battery. J Power Sources 2012;208:210–24. using ultrasonic transmission–Concept and laboratory testing. J Power Sources
[234] Schmidt JP, Arnold S, Loges A, et al. Measurement of the internal cell temperature 2017;343:536–44.
via impedance: evaluation and application of a new method. J Power Sources [263] Ladpli P, Kopsaftopoulos F, Chang FK. Estimating state of charge and health of
2013;243:110–7. lithium-ion batteries with guided waves using built-in piezoelectric sensors/
[235] Hu X, Feng F, Liu K, et al. State estimation for advanced battery management: key actuators. J Power Sources 2018;384:342–54.
challenges and future trends. Renew Sustain Energy Rev 2019;114:109334. [264] Kim Y, Samad NA, Oh KY, et al. Estimating state-of-charge imbalance of batteries
[236] Lee CY, Lee SJ, Hung YM, et al. Integrated microsensor for real-time microscopic using force measurements[C]. In: 2016 American control conference (ACC). IEEE;
monitoring of local temperature, voltage and current inside lithium ion battery. 2016. p. 1500–5.
Sensor Actuator Phys 2017;253:59–68. [265] Samad NA, Kim Y, Siegel JB, et al. Battery capacity fading estimation using a
[237] Zhang G, Cao L, Ge S, et al. In situ measurement of radial temperature force-based incremental capacity analysis. J Electrochem Soc 2016;163(8):
distributions in cylindrical li-ion cells. J Electrochem Soc 2014;161(10): A1584–94.
A1499–507. [266] Zou C, Manzie C, Ne�si�c D, et al. A framework for simplification of PDE-based
lithium-ion battery models. IEEE Trans Contr Syst Technol 2016;24(5):1594–609.
17
[267] Zou C, Zhang L, Hu X, et al. A review of fractional-order techniques applied to [271] Hu X, Jiang H, Feng F, Liu B. An enhanced multi-state estimation hierarchy for
lithium-ion batteries, lead-acid batteries, and supercapacitors. J Power Sources advanced lithium-ion battery management. Appl Energy 2020;257(1):110419.
2018;390:286–96. [272] Liu K, Zou C, Li K, et al. Charging pattern optimization for Lithium-ion batteries
[268] Hu Y, Yurkovich S, Guezennec Y, et al. Electro-thermal battery model with an electrothermal aging model. IEEE Trans Indus Inf 2018.;14(12):5463–74.
identification for automotive applications. J Power Sources 2011;196(1):449–57. [273] Liu K, Hu X, Yang Z, et al. Lithium-ion battery charging management considering
[269] Song Y, Liu D, Liao H, Peng Y. A hybrid statistical data-driven method for on-line economic costs of electrical energy loss and battery degradation. Energy Convers
joint state estimation of lithium-ion batteries. Appl Energy 2020;261:114408. Manag 2019;195:167–79.
[270] Feng F, Teng S, Liu K, et al. Co-estimation of lithium-ion battery state of charge [274] Tang X, Liu K, Wang X, et al. Real-time aging trajectory prediction using a base
and state of temperature based on a hybrid electrochemical-thermal-neural- model-oriented gradient-correction particle filter for Lithium-ion batteries.
network model. J Power Sources 2020;455:227935. J Power Sources 2019;440:227118.
18

10 1016@j Rser 2020 110015

Uploaded by

Copyright:

Available Formats

10 1016@j Rser 2020 110015

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

10 1016@j Rser 2020 110015

Uploaded by

Copyright:

Available Formats

Renewable and Sustainable Energy Reviews 131 (2020) 110015

Contents lists available at ScienceDirect

Renewable and Sustainable Energy Reviews

A comprehensive review of battery modeling and state estimation

1. Introduction obtained by direct measurement approach and is easily affected by

Fig. 1. Functional structure diagram of an advanced battery management system.

2.2.1. Integral-order equivalent circuit models

Fig. 2. Key features of the battery management system.

Fig. 4. Benefits and drawbacks of different electrical equivalent circuit models.

Fig. 5. The comparison of different SOC estimation approaches.

Fig. 6. Key benefits and drawbacks of different SOH/RUL estimation approaches.

Fig. 7. Advanced BMS architecture with 5G.

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.