Energy 183 (2019) 220e234

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

Towards a smarter battery management system: A critical review on

optimal charging methods of lithium ion batteries
Qian Lin a, Jun Wang a, Rui Xiong a, *, Weixiang Shen b, Hongwen He a
a
National Engineering Laboratory for Electric Vehicles, School of Mechanical Engineering, Beijing Institute of Technology, Beijing 100081, China
b
Faculty of Science, Engineering and Technology, Swinburne University of Technology, Hawthorn, Victoria, 3122, Australia

a r t i c l e i n f o a b s t r a c t

Article history: Automotive electrification is a main source of demand for lithium ion batteries. Performances of battery
Received 22 March 2019 charging directly affect consumers' recognition and acceptability of electric vehicles. Study on optimized
Received in revised form charging methods is vital for future development of a smarter battery management system and an
4 June 2019
intelligent electric vehicle. This paper starts from introducing the working principles and existing
Accepted 19 June 2019
Available online 20 June 2019
problems of simple charging methods and then elaborating various optimized charging methods along
with their characteristics and applications. It demonstrates that the optimized charging methods can
reduce charging time, improve charging performance and extend battery life cycle comparing with
Keywords:
Electric vehicle
conventional charging methods. At the end, this paper also provides a four-step pathway towards the
Lithium ion battery design of an optimal charging method of Li-ion batteries: determine optimization objectives, establish
Optimized charging methods optimization scheme, develop matching design and implement and promote the optimal charging
Charging time method.
Life cycle © 2019 Elsevier Ltd. All rights reserved.

1. Introduction constant-current (CC) or constant-voltage (CV) charging, which

both are inefficient and incapable of satisfying the growing demand
Recent years have witnessed a booming development of electric of modern EV industries [16, 17]. Various attempts have been made
vehicle (EV) industries [1, 2]. Although numerous commercial to develop improved approaches to charge Li-ion batteries through
batteries are used as energy storage systems to power EVs, lithium the accelerated charging process, the improved charging perfor-
ion (Li-ion) batteries have become one of the most popular battery mance or the maximized service life [18, 19]. These improved ap-
technologies in EVs due to their high energy and power density, proaches can be classified from different perspectives, such as
long life cycle and low rate of self-discharge [3e5]. However, wide charging waveform improvement, battery model coupling and the
deployment of Li-ion batteries in EVs has been greatly restricted by latest alternating current (AC) charging. These approaches have
their existing charging strategies [6, 7]. To address this issue, the delivered a significant performance boost in practice.
research on the provision of an optimal charging method to Li-ion The first category of improved charging methods is to transform
batteries has emerged as a new paradigm towards a smarter bat- either current waveform or voltage waveform or both to improve
tery management system (BMS) [8,9]. Li-ion battery systems in EVs charging profiles. Besides the common constant current-constant
consist of hundreds to thousands of cells in series or parallel. The voltage (CC-CV) method, there are multiple alternatives such as
development of an optimal charging method to a Li-ion battery multi-stage CC using integer linear programming (ILP) algorithm
system in EVs is closely associated with state of health (SOH) of [20], multi-stage CC-CV based on the evolution of battery internal
each cell [10e12] in the system, which poses considerable chal- resistance [21]. Moreover, pulse charging is also developed by using
lenges [13e15]. either pulse current or pulse voltage to achieve ions diffusion and
Simple charging methods of Li-ion batteries are either through neutralization during the rest period of the pulse. Pulse charging
has been proved that it can significantly reduce the time required
for charging and extending service life of lead-acid batteries, which
* Corresponding author. Department of Vehicle Engineering, School of Mechan- provides some guidelines for the development of more advanced
ical Engineering, Beijing Institute of Technology, No. 5 South Zhongguancun Street, charging methods for Li-ion batteries [22,23]. Waveform-based
Haidian District, Beijing 100081, China. charging methods have been widely explored to charge Li-ion
E-mail address: [email protected] (R. Xiong).

https://doi.org/10.1016/j.energy.2019.06.128
0360-5442/© 2019 Elsevier Ltd. All rights reserved.
 Q. Lin et al. / Energy 183 (2019) 220e234 221

batteries nowadays [24e26]. the optimized charging methods, which lays the foundation for a
The second category of the improved charging methods is to further development of a smarter BMS for EVs.
apply battery models such as equivalent circuit models (ECM) and The rest of this study is organized as follows. The second section
electrochemical models (EM) to optimize charging performances is to introduce simple charging methods. The third section details
[27,28]. In Ref. [29], an ECM based charging method with the model some of the optimized charging methods available, followed by a
parameters extracting from experimental data and a universal brief summary and comparison. Finally, the study provides the
voltage protocol (UVP) is able to improve charging efficiency and directions for the development of new charging methods. The
extend battery life cycles. In order to study the internal character- conclusions are given in the last section.
istics of a Li-ion battery during the charging process, charging
strategies based on EM are also proposed [30, 31]. Experimental 2. Simple charging methods
results show that the EM based charging methods are capable of
greatly shortening charging time while meeting the requirements Simple charging methods in this study refer to a charging
of health-related conditions. scheme which only one parameter either current or voltage re-
The charging methods discussed above fail to achieve satisfac- mains constant in a whole charging process. They are rarely seen in
tory results in battery systems with high performance and high modern BMSs for EVs. However, it is still worth starting our dis-
power requirements of charging equipment [32,33]. The solution is cussion on these simple charging methods to investigate their
to use AC charging, which leads to the third category of the drawbacks and the necessity of developing the optimized charging
improved charging methods [34e36]. This charging method has methods.
gradually drawn a wide attention. The recent research focuses on
finding an optimal charging frequency, which varies nonlinearly
2.1. Constant current charging
with temperature, state of charge (SOC), and SOH of a lithium ion
battery. In Ref. [37], an online tracking algorithm is developed to
Constant current (CC) charging requires maintaining constant
dynamically track an optimal charging frequency under any
current during the whole charging process, which is normally
conditions.
based upon an accurate prediction of SOC of the battery [38,39].
To identify the commonalities and advantages of the improved
This method has limited current to prevent over-current in an
charging methods in those three categories, a summary of a diverse
initial charging process [40], and also has the advantage of deter-
range of the improved approaches to charge Li-ion batteries based
mining charging current easily, which depends only upon battery
on constant current or voltage waveform, battery model and AC
capacity and charging time. The CC charging method is widely used
waveform are shown in Fig. 1. This helps to discover the evolution of
to charge nickel-cadmium batteries [41]. However, its main prob-
lem is that the cumulative errors in the existing SOC estimation
algorithms may lead to either overcharging or undercharging [42,
43], resulting in a reduction of battery life cycle. Hence, CC charging
methods are rarely adopted in modern BMSs for EVs.
Standard CC charging method is characterized by a high current,
usually between 0.2C and 1C. This makes it much less time-
consuming and easy to operate [44]. However, as the charging
process continues, the acceptable level of charging current gradu-
ally declines as shown in Fig. 2. If high charging current remains
unchanged at a later stage of charging, active materials on the
electrode plate will be detached, thereby causing battery capacity
losses and affecting battery service life [45].
Trickle charging is derived from CC charging method. It provides

Fig. 1. Summary of various charging methods. Fig. 2. Charging rate versus battery SOC.
222 Q. Lin et al. / Energy 183 (2019) 220e234

a continuous CC at a low rate (0.01Ce0.1C) to compensate self-

discharge rate and maintain fully charged condition. This method
is normally suitable for repairing and activating batteries and is
widely used for starting, lighting, and ignition (SLI) batteries ap-
plications [46].

2.2. Constant voltage charging

Constant voltage (CV) charging requires the voltage of power

supply to be constant throughout the charging process. During the
progress of CV charging, the terminal voltage of battery gradually
increases, which causes the gradual decrease in charging current.
The charging process ends once a preset lower limit of charging
current is reached, which can effectively avoid overcharging to
extend the service life of the batteries regardless of some irre-
versible negative effects inside the battery.
CV charging allows charging current to be adjusted based on the
estimated SOC and a properly-set level of constant voltage can Fig. 3. CC-CV charging.
ensure that Li-ion batteries are fully charged. However, CV charging
has its disadvantage in causing damage to batteries due to its
relatively larger charging current in an initial charging period with increase at a relatively fast pace during charging, thereby resulting
low SOCs. Such large charging current is far more than levels in noticeable polarization and higher polarization voltage. 2. the CV
suitable for battery charging, which causes crystal lattices to phase is very time-consuming, which is often considered unsuit-
collapse inside a battery with temperature rise. As can be seen from able for a quick charging process [48]. 3. a relatively high current
Fig. 2, charging current starts to linearly decrease with time in the tends to be adopted in the CC charging phase to reduce charging
low range of SOC while charging current remains low level for a time, which exacerbates side reactions inside the charged battery
long period in the high range of SOC, which sharply reduces and affects its cycle life. 4. the method is independent of battery
charging speed and increases charging time [47]. models so that it cannot distinguish between individual cells. 5. the
internal resistance of the batteries is neglected and a pre-
3. Optimized charging methods determined charging process likely causes a surge in battery tem-
perature and a drop in charging efficiency [49].
It is clear from the foregoing discussions that simple charging Another popular charging method is to add a trickle charging
methods for Li-ion batteries no longer satisfy the demand of stage before CC stage and an end-of-charge stage after CV stage in
modern EV industries. This leads to the development of various CC-CV charging, which has been seen in modern BMSs. It consists of
optimized charging methods, which are introduced in the four stages and is often referred to as trickle charge-constant cur-
following section. rent-constant voltage charging method (TC-CC-CV) [50] The first
stage, trickle charging, is activated only when the battery is deeply
3.1. Waveform-based charging methods discharged. The last stage, end-of-charge, is triggered when the
charging current in CV stage drops to a predefined threshold. This
Waveform-based charging methods refer to modifying current last stage is similar to trickle charging which completes charging at
or voltage waveforms based on CC charging, CV charging and CC-CV a significantly reduced charging current, resulting in a longer
charging to reduce charging time and improve charging charging time. As a result, this method greatly prolongs cycle life of
performance. Li-ion batteries [51].
Many suggestions have been given to improve the CC-CV
3.1.1. i. Constant current-constant voltage charging charging. For instance, in the charging process measured current
To fully take the advantages of CC charging and CV charging and voltage can be used to predict SOH and SOC. Based on the
while overcoming their drawbacks, the constant current-constant predicted SOC and SOH, the preset charging current or voltage can
voltage (CC-CV) charging method derived from the simple be timely adjusted to achieve a real-time control of charging pro-
charging methods of CC charging and CV charging is recommended. cess. Fig. 4 (a) and (b) shows two optimal CC-CV approaches with
It is the most common method to charge Li-ion batteries in modern either an improved CC phase [52] or an improved CV phase [53]. In
EVs at present. Many optimized charging methods have been Fig. 4 (a), a charging voltage slightly over the upper limit is applied
developed on the basis of a CC-CV charging method. It is charac- to accelerate charging phase before the start of CC charging. Ac-
terized by a preset current for charging at CC phase, when the cording to the experimental results, it delivered an optimized
voltage of charged batteries rises to a preset level it shifts to CV charging current distribution while reducing the decline in battery
charging. During CV phase, when the charging current reduces to capacity. In Fig. 4 (b), a grey-prediction (GP) algorithm was applied
cut-off current, the entire charging process is completed. Fig. 3 in the CV phase of the standard CC-CV method to predict the
shows the changes in current and voltage throughout the whole charging behavior periodically for determining the optimized
CC-CV charging process. Generally, CC charging is the main part of charging current while the charging current of the CC phase re-
the entire charging process for CC-CV charging, making up over 85% mains the same. This charging method is named as CC-GP charging,
of the rated capacity for Li-ion batteries. where the charging current of the CV phase is larger at higher
Currently, the CC-CV charging method is a major means of current part and smaller at the lower current part than that of the
charging Li-ion batteries as it has no need for the knowledge of CV phase of the standard CC-CV charging method. As a result, it
battery models. The CC-CV charging circuit is also easy to design, maintains the optimized charge trajectory to accelerate charging
implement and operate. However, it poses various issues: 1. with process and its implementation can be explained as follows: the
the decrease in battery capacity due to aging, battery voltage will measured voltage during CC charging was used to predict the level
 Q. Lin et al. / Energy 183 (2019) 220e234 223

Fig. 4. Improved CC-CV methods: (a) pre-accelerated charging method; (b) grey-predict charging method.

of voltage in the following cycle Ts to determine charging current as current, the upper and lower limits of cut-off voltage as well as the

 SOC interval. By solving the optimization problem in Eq. (2) under
4:2  vo ð4:2  ~vo Þ the constraints, the charging current of each stage can be obtained.
ic;gp ¼ min 0:7C; þb, (1)
rr rr The relevant charging curve is shown in Fig. 5. Y.H. Liu et al. per-
formed the charging of Li-ion batteries by applying a five-stage
where ic,gp is the charging current of the proposed GP charging current (2.1 C、1.7 C、1.5 C、1.3 C and 1.0 C) in Ref. [55]. In this
strategy, rr is the inter-pack-resistance, vo is the open circuit way, they can charge the battery to around 70% of the rated capacity
voltage, ~vo is the predicted open circuit voltage, b is the enhance- (930 mAh) in 30 min. As compared to the conventional CC-CV
ment factor. The improvement lies primarily in the CV phase. Ifb ¼ method, this charging method can extend battery cycle life by
0, this charging method is simplified to the standard CC-CV. The 25% with an attenuation ratio of 25%. The ant colony system algo-
CC-GP method is superior to the standard CC-CV in terms of the rithm (ACS) was used to obtain the optimized charging current for
charging speed and efficiency by 23% and 1.6%, respectively. It is each stage which is significantly affected by model parameters.
even more desirable compared with the CV charging method as it Furthermore, Y.H. Liu et al. applied the continuous orthogonal al-
increases the charging speed by 34% and boosts the charging effi- gorithm [25], which was based on the Taguchi method to deter-
ciency by 7%. Furthermore, the charging speed of the CC-CV can be mine the optimal five-stage current (1.45C、1.05C、1.00C、0.700C
effectively improved by increasing charging current during the CC and 0.10C) and was able to charge a 650 mAh battery to over 95%. As
phase or raising the upper limit of cut-off voltage during the CV compared to the conventional CC-CV applying a charging current of
phase [54]. 1.45C, the charging time is reduced by 11.2%, the charging efficiency
is improved by 1.02%, and the cycle life is extended by approxi-
3.1.2. ii. Multi-stage constant-current charging mately 57% with a 30% attenuation rate.
Multi-stage CC charging represents another solution to exces- In Ref. [56], the particle swarm optimization (PSO) algorithm
sively long charging time required in the CV phase of the CC-CV. To characterized by its fast convergence was also applied to the multi-
reduce the charging time, a high current is necessary for charging stage CC charging method to determine the optimized charging
process, causing the terminal voltage to hit the upper limit of cut- current for each stage. Its shifting condition based on the upper cut-
off voltage in a short time while failing to gain expected charging off voltage was used to optimize the charging current by simulating
capacity. This problem can be solved by multi-stage CC charging. It the behavior of PSO in 16 cells and determining the weights of
is implemented as follows. When the first preset current is applied charging time and charging capacity in the objective function by
to charge a battery until the battery voltage reaches to the upper fuzzy control method. The experimental results indicate that this
limit of cut-off voltage, the charging process shifts to the following multi-stage CC charging strategy can remarkably shorten the
preset current and repeats the previous charging process until all charging time by 56.8%, prolong the battery life by 21% and improve
the preset levels of charging current are used. There is a gradual the energy efficiency by about 0.4% compared with the conven-
decline in the preset charging current at each stage to prevent the tional CC-CV.
charged battery from reaching the upper limit of cut-off voltage too
quickly. The shifting condition can also be set based on the limit of
the SOC interval besides the upper cut-off voltage. The following
section gives an introduction to the charging methods for the SOC
and voltage based shifting conditions.

3.1.2.1. Shifting condition based on upper cut-off voltage. A multi-

target function is developed to calculate the level of current for
each stage through the following equation:

f1 ðiL Þ ¼ x1 ðiL Þ þ y1 ðiL Þ (2)

where iL is the charging current, x1, y1 are the functions of iL, which
represent different optimization objectives.
Since the limit of the upper cut-off voltage is used as a shifting
condition, the level of current for each stage is preset to decrease
incrementally before setting the optimal boundary conditions. The
boundaries can be set as a series of constraints like the maximum Fig. 5. Five-stage CC charging with shifting condition based on upper cut-off voltage.
224 Q. Lin et al. / Energy 183 (2019) 220e234

3.1.2.2. Shifting condition based on SOC interval. Similar to Eq. (2), denotes the concentration of Li-ions. Pulse charging methods can
the same form of multi-target function is established after setting be divided into two groups: current pulse charging and voltage
shifting condition based on the SOC interval to calculate the level of pulse charging.
current for each stage subject to a series of constraints. Subse-
quently, the limits of current for each stage are set. 3.1.3.1. Current pulse charging. Three different types of current
In Ref. [24], a four-stage CC charging method was proposed by pulse charging methods with constant-current and constant-
using the orthogonal array technique. The whole charging process frequency (CCCF-PC) are shown in Fig. 7. Experimental results
is equally divided into four stages in terms of the SOC interval as reveal that these three methods cannot significantly improve the
shown in Fig. 6, where the battery is charged by a preset current charging efficiency of the battery. Based on the analysis of CCCF-PC
(1.8 C, 1.3 C, 0.9 C and 0.5 C) in every 25% SOC interval. The charging charging methods, the current pulse charging with constant-
process will terminate when the estimated SOC reaches 100%, current and variable-frequency (CCVF-PC) and variable-current
which means the battery is fully charged. The experimental results and constant-frequency (VCCF-PC) are proposed, as shown in
verify that the four-stage constant CC method is more effective than Fig. 8. These two charging methods reduce polarization voltage by
CC-CV charging method in that a charging time is shorter than that giving rest time in the charging process to balance ion concentra-
of CC-CV (80 min) by 15 min or a reduction of 22.5% and the tem- tion and ensure that the internal state of the battery tends to be an
perature variation is almost half of that of CC-CV. Apparently, the equilibrium state, which helps to increase charging acceptance
SOC estimation is crucial in determining whether it should be while reducing the charging time.
shifted to the next charging stage. Therefore, the multi-stage
charging method, with the shifting condition based on SOC inter- 3.1.3.2. Voltage pulse charging. There are two voltage pulse
val, needs an accurate estimation of SOC in real time. Moreover, charging methods: one is based on duty-varied voltage pulse and
many other scholars have studied the multi-stage constant-current the other is based on variable-frequency voltage pulse.
charging methods to improve charging performances in terms of 3.1.3.2.1. Variable-duty voltage pulse charging. In Ref. [62], the
charging speed, charging efficiency and low energy loss [57e59]. duty-varied voltage pulse charger (DVVPC) was used to charge a Li-
ion battery. Fig. 9 depicts the corresponding time sequence dia-
3.1.3. Pulse charging gram. TF, TS and TC correspond to the full-charge detect mode
Pulse charging can be perceived as a discontinuous CC or CV (FCDM), the sense mode (SM) and the charge mode (CM),
charging (for current pulse and voltage pulse, respectively). It was respectively.
first employed in fast charging of lead acid batteries and then The working process is explained as follows: 1. a controller de-
explored to charge Li-ion batteries. The major characteristic of tects the charging state of the battery in FCDM and if it is fully
pulse charging is to eliminate or reduce polarization voltage to charged, the charging process is completed. 2. if the battery is not
allow acceptable current in the next cycle higher than that in other fully charged, the SM is to search for the suitable duty cycle Ds. 3.
charging methods. This will make the average charging current in after determining Ds, the voltage pulse with the suitable duty Ds is
pulse charging higher than that in other charging methods to produced in the CM to charge the battery during the charging
reduce charging time [60]. period TC. Consequently, a series of CV pulses at the constant fre-
In Ref. [61], pulse charging was to add a short-time rest interval quency was produced with different duty cycles Di (i ¼ 1, 2, 3, 4 …),
or a short-time discharging period during charging process to
reduce or eliminate polarization voltage in batteries. During
charging process, achieving the equilibrium of ion concentration
can improve charging efficiency because diffusion rate of lithium
ion is the fundamental reason of determining charging rate of a Li-
ion battery. The diffusion equation of Li-ions is written as:

vCLi ðx; tÞ v2 CLi ðx; tÞ

¼ DLi (3)
vt vx2

where DLi denotes the diffusion coefficient of Li-ions in solution, CLi

Fig. 6. Four-stage CC charging with shifting condition based on limit of SOC interval. Fig. 7. CCCF-PC methods.
 Q. Lin et al. / Energy 183 (2019) 220e234 225

frequencies. Then, it collects charging current at each frequency

and calculates the average value of the currents as

1 XM
Iagv ðnÞ ¼ I ðmÞ (4)
M m¼1 ib;n

where Iib,n denotes the mth sampling current value in the pulse
Fig. 8. (a) CCVF-PC method; (b) VCCF-PC method. charging at the frequency of fn, Iagv is the average current value of
the m sampling currents.
This charging method can dynamically detect and track the
optimal charging frequency by observing the charging response at
variable frequency pulses. Thus, the maximum average current
value can be determined by the optimal charging frequency
    
foptimal ¼ fn MAX Iavg ðnÞ ; n ¼ 1…N (5)

To assess the performance of variable-frequency pulse charging,

its prototype was designed and implemented to charge a 600 mAh
Li-ion battery. Comparing with the standard CC-CV charging and
fixed-frequency pulse charging, the variable-frequency pulse
charging increases charging speed by more than 24% and 10%,
respectively. In practice, it is recommended that the proper ratio of
charging time to searching time should be set to achieve better
charging performance and the thermal sensor and fault diagnosis
sensor should be used for safety reasons when the charging current
of the designed variable-frequency pulse charging is greater than
1C [64].
The pulse charging method has a strict requirement of the
Fig. 9. Time sequence diagram of DVVPC charging [62]. charging current since the magnitude of current is closely associ-
ated with the formation of solid electrolyte interface (SEI) film.
the corresponding average current values could be used to find the Under high current magnitude, it will form thick SEI film. The
optimal Di values. thickness of the SEI film exhibits an inseparable relationship with
Fig. 10 demonstrates the suitable duty Ds in voltage pulse battery performance because the increase in the thickness of the
charging based on the DVVPC during a charging process. It is SEI film causes the increase in battery impedance. Furthermore,
evident that traditional duty-fixed voltage pulse charging is not excessive high current will make Li-ions in a battery to generate
perfectly suitable. To assess the performances of this charging considerable crystal accumulation near the terminal post in the
method, duty-varied voltage pulse charging was compared with final charging phase, bringing about irreversible damage to the
CC-CV and duty-fixed voltage pulse charging. The experimental battery [65]. The comparison between pulse charging and CC-CV
results show that the charging speed and charging efficiency of the charging [66, 67], suggests that in the charging process of using
DVVPC are 14% and 3.4% better than CC-CV charging. Compared the same average current value, with Q ¼ I2*R, it could be found that
with duty-fixed voltage pulse charging (D ¼ 50%), the charging the current quadratic mean of the pulse charging is relatively high.
speed increases by about 5% and the charging efficiency increases Accordingly, more heat will be generated to cause quicker rise in
by approximately 1.5%. Thus, the duty-varied voltage pulse battery temperature in pulse charging than in CC-CV charging. Such
charging method with the suitable duty can provide the better phenomenon will become obvious with the increase in charging
charging performance. current amplitude, thereby accelerating the aging of a battery.
3.1.3.2.2. Variable-frequency voltage pulse charging. Hence, it is generally believed that the pulse charging method is
Variable-frequency voltage pulse charging methods was investi- inferior to the CC-CV charging method.
gated in Ref. [63]. It first charges batteries with voltages at various
3.1.4. Boost charging
With wide applications of Li-ion batteries, fast charging of Li-ion
batteries is necessary under many circumstances, i.e. vehicles with
passengers in emergency [50]. Boost charging was first proposed by
Notten et al. It can charge a fully discharged battery to one-third of
its rated capacity within 5 min [68].
As we mentioned earlier, simple CV charging method starts with
extremely high current, which reduces charging time but causes a
severe impact on battery life. For example, CV charging method
reduces charging time by 60% comparing with CC-CV charging
method. However, the former can cause capacity loss by 40% after
160 cycles whereas the latter only reduce capacity by 15% after 300
cycles. The concept of boost charging is to apply very high current
for a very short time to charge a fully depleted battery and then
switch to standard CC-CV charging method. In this way, boost
Fig. 10. Suitable duty in voltage pulse charging based on DVVPC during a charging charging will not introduce any negative degradation effects on Li-
process [62]. ion batteries. Experimental results demonstrate that both
226 Q. Lin et al. / Energy 183 (2019) 220e234

cylindrical and prismatic Li-ion batteries have similar cycle life (i.e. where the sampling interval Dt is set to 1s. Then, the multi-target
up to 700 cycles) when they both are charged by boost charging function of charging time and temperature rise is used to estab-
and standard CC-CV charging. lish the fitness function as

3.2. Model-based charging methods ð1  bÞtðiL Þ bTc ðiL Þ

FðiL Þ ¼ þ (9)
a b
Model-based charging methods calculate the optimal charging
where b denotes the weight coefficient of charging temperature
current of a battery using ECM or EM [69,70]. They achieve the goal
rise, t is the charging time, Tc is the temperature rise, a and b are the
of optimizing charging process by controlling polarization voltage.
coefficients for normalization.
The optimization of charging process based on genetic algo-
3.2.1. i. Equivalent circuit model-based charging
rithm is shown in Fig. 12 and it is performed and completed by the
3.2.1.1. Equivalent circuit/temperature model. In the ECM based genetic algorithm toolbox in MATLAB.
charging method, an ECM needs to be first established for a battery Also, the charging current curves of the whole charging process
and the Thevenin model shown in Fig. 11 is taken as the example can be calculated according to the heat production and constraint
[71]. In Ref. [72], the model parameters are assumed to be constant conditions:
within unit sampling time, the polarization voltage and the ter-
vffiffiffiffiffiffiffiffiffiffiffiffi
minal voltage can be discretized as: u
uJ  U 2D
8 t RD
<U iL ¼ (10)
D;k ¼ expð  Dt=tÞUD;k1 þ RD iL;k1 ½1  expð  Dt=tÞ Ri
(6)
: Ut;k ¼ Uoc;k þ UD;k þ iL;k Ri
where J denotes the heat production rate, and iL is the charging
current.
where Dt denotes the sampling interval, t denotes a time constant The constraints of the optimization problem include the tar-
and equals CD*RD, CD and RD are the diffusion capacitor and diffu- geted SOC in the charging process, the SOC at the end of charging,
sion resistance, iL is the load current, Uoc is the open circuit voltage, the upper limit of battery temperature and the acceptable
Ri is the internal ohmic resistance. maximum current during charging process. According to the
Battery temperature model is also established based on Fourier experimental results, the charging time and the temperature rise of
thermodynamic formula as follows [73]: the equivalent circuit/temperature model-based charging strategy
8 can be reduced by 7.89% and 38.8%, respectively, compared with
>
>
>
> those of CC-CV charging, which proves that the optimization im-
< mC vT ¼ Q þ Q þ Q
vt S O proves the charging performance of the model-based method.
(7)
>
>
> Q ¼ hAðT  Tamb Þ
>
:
3.2.1.2. Equivalent circuit/power loss model. In Ref. [74], a power
loss model was built to describe power loss of a Li-ion battery. The
where m denotes the unit quality, C is the battery specific heat minimum energy-loss is selected as an optimization objective
capacity, T is the battery temperature, t is the time, Qs is the subject to the constraints of the maximum charging time, current,
reversible reaction heat rate, Qo is the power loss rate, Q is the heat voltage and charging quantity of electricity. The optimization
dissipation rate, h is the convection coefficient and A is the battery function is highly nonlinear and time varying. The optimization
surface area, Tamb is the environmental temperature, the total heat toolbox in MATLAB was employed to simulate and optimize the
production rate of the battery is the sum of Qo and Qs. charging current of the Li-ion battery. Subsequently, the Optimal
Discretizing eq. (7) obtains battery temperature at the time of k Current Profile (OCP) was identified.
as According to Ref. [74], for level I charging, CC-CV is nearly the



optimal charging method for minimizing power loss when the
hA hA Qo;k þ Qs þ hATamb
Tk ¼ exp  Tk1 þ 1  exp  temperature is assumed to be constant. Under the same condition
mC mC hA of charging capacity and charging time, the charging efficiency of
(8) the OCP was improved by only 0.42% compared with that of CC-CV.
Given the relatively small difference of power loss and the simi-
larity between OCP and CC-CV curve, CC-CV charging can be
considered as the approximation of power loss minimization.
However, for levelIIcharging where high charging currents are
used, battery temperature control to minimize power loss can be
realized with the temperature as a state variable to be built into the
power loss model. In this case, the optimal charging profile is found
to be quite different from CC-CV. The cost function for power loss
minimization is expressed as

  2 ðtf
J ¼ x1 tf  SOC* þ Ploss dt (11)
0

where SOC* denotes the targeted SOC, Ploss denotes the power loss.
Through adjusting the weights of the parameters in the cost
functions, the optimal charging strategy can identify charging
Fig. 11. Thevenin equivalent circuit model. profiles with different requirements or limits, such as the
 Q. Lin et al. / Energy 183 (2019) 220e234 227

Fig. 12. The optimization process basing on genetic algorithm [72].

temperature and SOC at the end of charging and total energy loss.
The experimental results demonstrate that the energy and power
density of the Li-ion battery are increased due to the initial rise of
battery temperature, particularly for cold climate. This results in
fuel savings and battery stress reduction.
This charging strategy stores the information of the best
charging current curves under different charging times and initial
SOC conditions into a micro-controller in advance, which can be
applied to battery heating in cold climates, thereby taking over the
part of responsibility of battery heating systems. Since it increases
battery temperature in the initial charging stage, it has significant
energy-saving effect in cold weather when new energy vehicles are
driven soon after the end of the charging process, and it also has Fig. 13. The block diagram of electrical-thermal-aging model [75].
major impact on how new energy vehicles are charged and utilized.

represents the weights of the relative importance between

3.2.1.3. Electrical-thermal-aging model. The two aforementioned charging time and capacity loss.
ECM-based charging methods have proven to be effective in com- It is easy to find that Eq. (12) combines two objectives, namely
parison with conventional charging methods. But, it is obvious that time consumption and capacity loss. Generally, this highly
the above methods only consider electrical behaviors of a battery nonlinear optimization problem based on the electrical-thermal-
while ignoring thermal and aging effects on battery charging. aging model is difficult to solve by conventional optimization
Taking these effects into account, a health-conscious fast and safe techniques. Pseudo-spectral methods were applied to overcome
charging protocol was developed by formulating a multi-objective this difficulty.
optimal control problem via a coupled electrical-thermal-aging
model [75]. This coupled model was integrated into the charging
control to optimize the entire charging process. Fig. 13 shows the 3.2.2. ii. Electrochemical model-based charging
block diagram of electrical-thermal-aging model, where I (t) is the The ECM based charging methods and waveform-based
charging current; Vt is the terminal voltage; Qt is the heat genera- charging methods cannot take into account internal chemical re-
tion; Tf, Ts and Tc correspond to the ambient temperature, the sur- actions, internal potential change and Li-ions concentration. This
face temperature and the core temperature, respectively. The can be solved by optimizing the optimum charging strategy based
objective function (J) for the entire charging process can be on the EM [76]. Based on the study of the EM of a Li-ion battery in
expressed by Ref. [77, 78], a one-dimensional EM was established in Ref. [79]. The
model covers various internal chemical and physical parameters
tf  t0 h  i during the charging process and distinguishes the difference be-
J¼b þ ð1  bÞ SOHðt0 Þ  SOH tf ; 0 < b < 1 (12)
tmax  t0 tween these parameters in different positions in the one-
dimension [80].
where t0 and tf denote the initial and final charging times, b Especially when the charging current is very large, some side
228 Q. Lin et al. / Energy 183 (2019) 220e234

reactions will occur inside a Li-ion battery which can cause battery battery. Then, an optimal charging profile was achieved by maxi-
temperature to rise rapidly [81]. So, the EM based charging method mizing the stored charge within a preset time while maintaining
focuses on the use of as large charging current as possible on the intercalation induced stresses below a threshold.
premise of restraining side reactions which can initiate the pre- Two scenarios were considered in their study: (1) charging a
cipitation of lithium inside a battery. The EM helps to accurately fully depleted battery for 1800 s; and (2) charging a half-depleted
determine the constraints, such as the maximum current and battery for 900 s. The optimized charging profiles were compared
temperature during the charging process, and its objective function with standard CC-CV methods at various charging rates (4C, 3C and
of the optimization is to shorten charging time. Mathematically, it is 2C). In both scenarios, the optimized charging method can store
expressed by more charge at the same intercalation induced stress, particularly it
can store 4% more charge than the CC-CV method at 2C charging
ðtf rate. However, these results are limited by isothermal assumption,
Iopt ðtÞ ¼ argminIðtÞ 1dt (13) which needs to be addressed in the future by using a thermal
0 model.

The condition to initiate side reaction is:

3.3. AC charging method
hsr ðx; tÞ ¼ Fs ðx; tÞ  Fe ðx; tÞ  Usr ðx; tÞ > 0 (14)
AC charging method, more specifically the sinusoidal-ripple-
and the constraints are: current (SRC) method, has recently attracted many attentions in
Li-ion batteries. Unlike pulse charging technique, SRC utilizes a si-
8 0  IðtÞ  Imax nusoidal current superimposed with a direct current (DC) as
>
> ðtf
< charging current [34]. Such method seeks to charge Li-ion batteries
Q ¼ IðtÞdt (15) at an optimal frequency obtained from electrochemical impedance
>
>
: 0 spectroscopy (EIS) analysis, and thus to minimize frequency-
TðtÞ  Tmax dependent impedance. As shown in Fig. 15(a), the ac-impedance
model of a Li-ion-battery consists of a charge transfer Rct, a War-
where Iopt(t) is the optimal charging current which ensures the burg impedance Zw, a double-layer capacitance Cd, an ohmic
shortest charging time; I(t) is the charging current; and hsr(x,t) is resistance Ro, and an anode inductance Ld.
the over-potential, when it becomes negative, side reactions will Warburg impedance Zw is resulted from concentration polari-
occur at negative electrode. To restrain side reactions, this value zation, is usually modeled by:
must maintain the positive during charging process. Tmax is the
maximum temperature rise; Фs(x,t) is the electrode potential; s s
Zw ¼ pffiffiffiffi  j pffiffiffiffi (16)
Фe(x,t) is the electrolyte potential; Usr is the battery voltage. u u
Nonlinear model predictive control (NMPC) was applied to solve
the optimization problem subject to the constraints of temperature where s is the Warburg coefficient and u is the operating angular
and potential imbalance ranges [82]. In Ref. [83], the control vector frequency.
parameterization (CVP) was applied to optimize the charging pro- The equation above reveals that Warburg impedance Zw can be
cess by monitoring the concentration change of chemical sub- ignored at low frequency (i.e. below 1 Hz). Consequently, the ac-
stances inside a battery to improve charging capacity. Two impedance model can be simplified, as shown in Fig. 15(b).
assumptions were made for this method: 1. when the terminal Supposing frequency of charging current is fs (or ws), thus the AC
voltage of the battery reaches the upper limit of cut-off voltage, the impedance of the battery is expressed as:
charging process ends. 2. the stored energy of the battery equals the 2 3 2 3
charge flowing into the battery which can be calculated by inte- 6 Rct
7 6 R2ct 7
grating the charging current. 6 ðws Cd Þ2 7 6 ws Cd 7
Zbattery ðws Þ ¼ 6R0 þ
2 7 þ j6ws Ld 
2 7
To verify the effectiveness of the proposed charging method, the 4 5 4 5
R2ct þ 1 R2ct þ 1
EM based dynamic optimization charging method is compared w s Cd ws Cd
with CC at 1C rate and CC at an optimized rate. The charging current (17)
(expressed by current density) and stored energy of the Li-ion
battery using the three charging methods are shown in Fig. 14. where Zbattery denotes a frequency-dependent impedance, which is
The current density of CC with optimized current is 18 A/m2, and minimized at an optimum operating frequency that can be subse-
the corresponding charging current density of CC with 1C is 30 A/ quently used in SRC charging.
m2. Fig. 14(b) is the curves of the stored energy corresponding to The optimum frequency fZmin can be derived from Eq. (17) as:
currents of the three charging methods. It shows the current curve
of EM based dynamic charging method demonstrates a non-linear 1 pffiffiffiffiffiffiffiffiffiffiffi
fZmin ¼ k1 (18)
attenuation, and the charging current is relatively smaller at the 2pRct Cd
end of the charging process compared with those of CC methods. pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2R0 R3ct C 2d þ2Ld R2ct Cd þR4ct C 2d
Therefore, the Li-ion battery will reach the upper cut-off voltage for where k ¼ Ld .
relatively long period, thereby charging more energy. This reveals The charging current, voltage, capacity and temperature con-
that EM based dynamic charging method is superior to simple cerning the SRC charging and pulse ripple-current charging (PRC)
charging methods in terms of charging time and stored energy. with the optimal frequency (998 Hz) and CC-CV charging are
Other than monitoring concentration change of chemical sub- shown in Fig. 16, respectively. Among these three methods, the SRC
stances, more recent work by Suthar et al. studied intercalation charging has the best performances in terms of charging rate,
induced stresses, which is one of the dominant mechanisms ac- charging efficiency and temperature rise. The charging time,
counting for capacity loss [84]. In their study, Suthar et al. adopted charging efficiency, maximum temperature rise and service life of
an isothermal pseudo 2-dimensional (P2D) model to incorporate the SRC charging are improved by 17%, 2%, 45.8% and 16.1%,
electrochemical, transport and thermodynamics processes inside a respectively, compared with those of CC-CV charging. They are also
 Q. Lin et al. / Energy 183 (2019) 220e234 229

Fig. 14. EM based dynamic optimization charging compared with other methods: (a) current density; (b) stored energy [83].

Fig. 15. (a) Complete Li-ion impedance model and (b) simplified model.
230 Q. Lin et al. / Energy 183 (2019) 220e234

Fig. 17. Electrical second-order RC battery model [90].

improved by approximately 0.24%, 0.27% and 16.47%, respectively,

compared with those of pulse charging. However, a very low
charging efficiency of only 97% for CC-CV charging method is used
for comparison whereas the typical charging efficiency for CC-CV
charging method has been reported above 99% in Ref. [6] [50],.
Such low efficiency poses the problem for the validity of their
results.
A few other publications also agree with that the SRC charging
method increases charging efficiency and reduce charging times
[85e87]. However, the improved performances of the SRC have
been recently challenged by Ref. [88e90]. In Ref. [88, 89], the re-
sults show that such improved performances are not observed on
prismatic Li-ion cells. In Ref. [90], a further battery impedance
analysis has been conducted, the same circuit-based model as
shown in Fig. 15(a) is used to prove that battery impedance is not
minimal at the optimum frequency fZmin when considering DC
components in SRC charging. As a result, the performances of the
SRC charging method are not significantly different from those of
the CC-CV charging method.
Previously, Warburg impedance Zw has been neglected to
simplify the complete circuit model in Fig. 15(a) into the simple
model in Fig. 15(b). This works only when the device operates
under AC current with a frequency greater than 1 Hz, but this fails
when the DC components of the SRC charging method is consid-
ered. In fact, as indicated by Eq. (16), Warburg impedance Zw be-
comes significantly greater than Rct, but much less than Zd at low
frequency limit. It's trivial to derive out that overall impedance
becomes:

s s
Zbattery ðuÞ ¼ Ro þ Rct þ pffiffiffiffi  j pffiffiffiffi (19)
u u
Eq. (19) reveals a completely different battery impedance from
Eq. (16), indicating the limits of using the optimum frequency fZmin
predicted by Eq. (18) for minimized impedance when considering
DC components. Furthermore, Eq. (19) also shows infinite battery
impedance at zero frequency. This requires proposing a different
equivalent circuit model to understand battery impedance in the
SRC charging method.
An electrical second-order RC battery model is proposed to
study the battery impedance with DC component (as shown in
Fig. 17) [90]. This model consists of a series connection of a resis-
tance Ra and two RC parallel circuits. The model component values
are obtained by fitting measured voltage and current waveforms at
various frequencies in the SRC charging and they will be used to
simulate the process of the SRC charging.

Fig. 16. Comparison of sinusoidal AC charging at 998 Hz with pulse charging at 998 Hz
and CC-CV charging: (a) current; (b) voltage; (c) capacity; (d) temperature [34].
 Q. Lin et al. / Energy 183 (2019) 220e234 231

Fig. 18. (Left panel) charging currents and (right panel) charging amounts of (a) (d) CC (0.5 C)-CV charging, (b) (e) SRC (0.5 C)-CV charging, and (c) (f) PRC (0.5 C)-CV charging [90].

The electrical second-order RC battery model directly dispar- rise. Therefore, theoretical analysis and experimental work are still
ages earlier argument that the SRC charging method seeks to required to further understand SRC charging before it can be
operate at an optimum frequency to minimize battery impedance. adopted by industry.
As we can see in Fig. 17, no matter which frequency that the SRC is
operating at, DC component always passes through the three re- 4. Discussions and suggestions
sistances, Ra, Rb and Rc. Since the battery is charged by DC
component instead of AC components, the real battery impedance It can be summarized that various charging methods are pro-
that determines charging time and charging amount is frequency- posed to attain tradeoffs between some optimization goals
independent. In fact, it is always the same as CC charging, and thus including charging time, temperature rising rate, charging effi-
the SRC charging method cannot minimize battery impedance. ciency (or minimum power loss) and battery life cycle. To find the
Fig. 18 shows the curves of charging currents and charging most desirable charging method for a particular application, it is
amounts for three charging methods. Obviously, these three necessary to compare some crucial characteristics of these charging
charging methods offer almost identical charging profiles. The methods. Table 1 provides their qualitative comparison.
charging time before CV phase for the CC, SRC and PRC is 5406 s,
5434 s and 5496 s, respectively, with an overall charging time of
4.1. Their comparison leads to the discussions
10404 s, 10244 s and 10340 s. The difference of charging time be-
tween the CC and the SRC is only 2%. Similarly, the difference of
3. The standard CC-CV charging method is a widely used charging
charging amount is also less than 2%. Thus, the SRC doesn't provide
strategy. Model-based charging methods can be flexibly com-
obvious improvements on charging performances. This small dif-
bined with other models according to specific optimization
ference can be attributed to different maximum temperature rise.
objectives. When the temperature rise rate needs to be strictly
Although the DC components among three charging methods are
controlled, the optimal current distribution can be solved by
the same, their RMS currents are significantly different. The RMS
combining ECM with a temperature model. Likewise, when the
currents of the SRC and the PRC are 22% and 41% greater than the
CC, respectively. As a result, the maximum temperature rises for
both the SRC (Tmax-rising ¼ 5.9  C) and the PRC (Tmax-rising ¼ 7.6  C) Table 1
are higher than the maximum temperature rise for the CC (Tmax- Characteristics comparison of charging methods.

rising ¼ 5 C). The extra heating of battery from the SRC might be the Charging methods Time Efficiency Complexity Cycle life
only reason for slightly improved charge acceptance which is not
CC M L L L
considered in Ref. [34]. Furthermore, a slightly high DC voltage in
CV H L L L
the CC-CV charging method can outperform the SRC-CV charging CC-CV M M M M
method [90]. Multi-stage CC L M M H
So far, there is no consensus on whether SRC-CV charging Current-PC M M M L
methods can improve charging performances of Li-ion batteries or Voltage-PC M M H L
ECM-based L/M H H/M H
not. Even, there are the conflict experimental results in tempera- EM-based L H H H
ture rise during SRC charging. The results in Ref. [34] show low AC M M/H H H
temperature rise but those in Refs. [89,90] show high temperature
Note that H, M or L stands for high, medium, and low, respectively.
232 Q. Lin et al. / Energy 183 (2019) 220e234

Fig. 19. Suggestions for designing optimal charging methods of a Li-ion battery.

power loses needs to be minimized, a power loss model will be define the operation details of each stage; d. identify the process
combined with ECM to provide a clever solution to the problem constraints. The key control points (KCPs) need to be set to ensure
with the minimum power loss as the optimization target. Future safety during the charging process. These KCPs can be voltage (rise)
work can develop SOX-dependent ECM. The accurate SOX esti- range, current range and temperature (rise) range. 3. develop
mation allows optimal control of battery charging process. matching design. The implementation of an optimal charging
4. The charging strategy considering aging model has attracted strategy in the BMS requires well-designed hardware and software,
many attentions and achieved better charging performance. The data acquisition, high computation capability and suitable control
impact of battery degradation on electrical parameters and and management system. 4. integrate the optimized charging
hence charging current pattern can be considered in the future method into the intelligent BMS. This intelligent BMS will work
study. with big data and the improved infrastructures such as efficient
5. EM-based charging methods have high requirement for charging piles and smart grid to realize cloud computing, fast
computational cost, charging equipment and data acquisition. charging, and even allocating energy resources, thereby building a
Waveform-based charging methods based on simplified models smart city.
or preset energy input can be adopted when the precision of
battery parameters is not strictly requested. In addition, various
optimization algorithms have been employed to obtain the 5. Conclusions
optimized charging current for the multi-stage charging
methods, which significantly reduces the charging time and This paper systematically reviews and compares the current
extend the cycle life of the battery. Future work is still necessary optimized charging methods for a Li-ion battery, which have ach-
to reduce computational cost as well as to include experimental ieved excellent performances in saving charging time, improving
validation of EM-based model. charging efficiency and extending battery life. This research work
6. Pulse charging and sinusoidal AC charging with an optimal also lays the foundation to development trend of a new optimized
charging frequency may be suitable for charging a large capacity charging strategy.
and high voltage battery system. However, the effectiveness of
those charging methods for some Li-ion batteries has been 1. Change from constant to variable charging pattern. On the basis
challenged recently, thus more statistical validations are of CC, CV and CC-CV, many optimized charging methods such as
required in the future study. improved CC-CV charging, multi-stage CC charging, pulse
charging are developed to charge a Li-ion battery. The research
Fig. 19 shows the suggestions towards the design of an optimal on the transition from DC charging to AC charging has also been
charging method of a Li-ion battery. carried out to charge Li-ion battery. These optimized charging
It can be divided into four steps: 1. determine the optimization methods have improved charging performances and provided
objectives, which may consist of charging time, charging efficiency, valuable experiences for further study.
life cycle and temperature rise. When a battery pack is charged, the 2. Adopt optimization and adaptive estimation algorithms. The
consistency of each cell in the pack should be maintained to avoid optimization and estimation algorithms should be applied in the
overcharge. 2. establish optimization scheme. In this step, there are design process of a charging strategy for a Li-ion battery [92]. It
four parts: a. decide the charging power input, which can be con- can find an optimal charging current and charging frequency
stant current or constant voltage, pulse current or pulse voltage, subject to the constraints, or calculate charging voltage at the
variable current or voltage with low frequency and AC current; b. next moment by using battery model parameters [93, 94]. Since
identify the process boundary to terminate a charging process, such battery model parameters are not constant due to battery aging,
as the upper limit of cut-off voltage, the lower limit of cut-off adaptive estimation strategy should be introduced to obtain
current and 100% SOC; c. design an optimal charging method in battery model parameters.
terms of the key success factors (KSFs) of the process workflow, 3. Apply an internal charging mechanism to charging process. The
these KSFs are advanced optimization algorithms, battery models internal charging mechanism of the Li-ion battery is closely
and equipment [91]. A complete process may break into stages and related to its chemical reaction mechanism inside the battery.
Studies reveal that chemical reaction processes play an essential
 Q. Lin et al. / Energy 183 (2019) 220e234 233

role in the charging process, such as internal potential and the [20] Dung LR, Yen JH. ILP-based algorithm for Lithium ion battery charging profile.
In: IEEE International Symposium on Industrial Electronics; 2010.
polarization of the battery and the change of lithium ion con- n D, Gonzalez M, Viera J, García V, Blanco C, Valledor M. Fast charging
[21] Ansea
centration [95e97]. An internal charging mechanism has been technique for high power lithium iron phosphate batteries: a cycle life anal-
applied in EM based charging and thermal model-based ysis. J Power Sources 2013;239:9e15.
charging, which have achieved excellent charging [22] Bhatt M, Hurley WG, Wolfle WH. A new approach to intermittent charging of
valve-regulated lead-acid batteries in standby applications. IEEE Trans Ind
performances. Electron 2005;52(5):1337e42.
[23] Li J, Murphy E, Winnick J, Kohl PA. The effects of pulse charging on cycling
The optimized charging strategies for lithium ion batteries in characteristics of commercial lithium ion batteries. J Power Sources
2001;102(1):302e9.
new energy vehicles have been developed for many years [98]. [24] Vo T, Chen X, Shen WX, Kapoor A. New charging strategy for lithium ion
With the rapid development of new energy vehicles recently, batteries based on the integration of Taguchi method and state of charge
intelligent BMS integrated with optimized charging strategies can estimation. J Power Sources 2015;273:413e22.
[25] Liu YH, Hsieh CH, Luo Y. Search for an optimal five-step charging pattern for
interact with smart grid, which may significantly affect the con- Li-ion batteries using consecutive orthogonal arrays. IEEE Trans Energy
struction of a smart city. The optimized charging method will Convers 2011;26(2):654e61.
become one of the research focuses in the field of new energy ve- [26] Luo Y, Liu Y, Wang S. Search for an optimal multistage charging pattern for
lithium ion batteries using the Taguchi approach. In: TENCON 2009 - 2009
hicles and will create a profound influence on shaping the future
IEEE region 10 conference. IEEE; 2009.
transportation. [27] Legrand N, Knosp B, Desprez P, Lapicque F, Ra L. Physical characterization of
the charging process of a Li-ion battery and prediction of Li plating by elec-
trochemical modelling. J Power Sources 2014;245:208e16.
Acknowledgement [28] Huang SJ, Huang BG, Pai FS. Fast charge strategy based on the characterization
and evaluation of LiFePO batteries. IEEE Trans Power Electron 2013;28(4):
This work was supported by the National Key Research and 1555e62.
[29] Guo Z, Liaw BY, Qiu X, Gao L, Zhang C. Optimal charging method for lithium
Development Program of China (Grant No.2017YFB0103802) and
ion batteries using a universal voltage method accommodating aging. J Power
the National Natural Science Foundation of China (Grant No. Sources 2015;274:957e64.
51877009). [30] Zou C, Hu X, Wei Z, Wik T, Egardt B. Electrochemical estimation and control
for lithium ion battery health-aware fast charging. IEEE Trans Ind Electron
2018;65(8):6635e45.
References [31] Chu Z, Feng X, Lu L, Li J, Han X, Ouyang M. Non-destructive fast charging al-
gorithm of lithium ion batteries based on the control-oriented electro-
[1] Wang Y, Sperling D, Tal G, Fang H. China's electric car surge. Energy Policy chemical model. Appl Energy 2017;204:1240e50.
2017;102:486e90. [32] Wen F, Jiang J, Zhang W. Charging method for Li-ion battery pack in electric
[2] Vaughan Adam. All Volvo cars to be electric or hybrid from 2019[EB/OL]. vehicles. Automot Eng 2008;30(9):792e5.
https://www.theguardian.com/business/2017/jul/05/volvo-cars-electric- [33] Leitman S. Modern electric, hybrid electric, and fuel cell vehicles. CRC press;
hybrid-2019; 2017. 2018.
[3] Yang G, Zhou Q, Hou M. Analysis of China's electric vehicles and its prospect. [34] Chen L, Wu S, Shieh D, Chen T. Sinusoidal-ripple-current charging strategy
Int Pet Econ 2017;25(4):59e65. and optimal charging frequency study for Li-ion batteries. IEEE Trans Ind
[4] Zheng Z, Zhang Z, Tang T. Recent development of lithium ion batteries. Chem Electron 2013;60(1):88e97.
World 2004;45(5):270e3. [35] Partovi Afshin. System and method for charging or powering devices, such as
[5] Hannan MA, Lipu MS, Hussain A, Mohamedb A. A review of the lithium ion robots, electric vehicles, or other mobile devices or equipment: U.S. Patent:
battery state of charge estimation and management system in electric vehicle No. 9,722,447, 2017.
applications: challenges and recommendations. Renew Sustain Energy Rev [36] Hare, William M, et al. Dynamic power limit adjustment in a battery charging
2017;78:834e54. process: U.S. Patent: 10,065,526, 2018.
[6] Shen WX, Thanh Tu vo, Kapoor Ajay. Charging algorithms of lithium-ion [37] Hussein Ala A, Abbas A, Samantha S. An online frequency tracking algorithm
batteries: an overview. 7th IEEE conference on industrial electronics and using terminal voltage spectroscopy for battery optimal charging. IEEE Trans
applications. ICIEA); 2012. Sustain Energy 2016;7(1):32e40.
[7] Sun Q. Electric vehicles: the breakthrough of "core" is the key - talking about [38] Chen X, Shen WX, Cao Z, Kapoor A. Adaptive gain sliding mode observer for
three bottlenecks in the development of the Li-ion battery. Adv Mater Ind state of charge estimation based on combined battery equivalent circuit
2010;12:62e3. model. Comput Chem Eng 2014;64:114e23.
[8] Zou C, Hu X, Wei Z, Tang X. Electrothermal dynamics-conscious lithium ion [39] Chen X, Shen WX, Cao Z, Kapoor A. A novel approach for state of charge
battery cell-level charging management via state-monitored predictive con- estimation based on adaptive switching gain sliding mode observer in electric
trol. Energy 2017;141:250e9. vehicles. J Power Sources 2014;246(3):667e78.
[9] Liu K, Li K, Yang Z, Zhang C, Deng J. An advanced lithium ion battery optimal [40] A. C. Hua and B. Z. Syue. Charge and discharge characteristics of lead-acid
charging strategy based on a coupled thermoelectric model. Electrochim Acta battery and LiFePO4 battery. The 2010 international power electronics con-
2017;225:330e44. ference (ECCE ASIA) 2010 1478-1483.
[10] Shahriari M, Farrokhi M. Online state-of-health estimation of VRLA batteries [41] J. G. Konopka, Constant current source and battery charger: U.S. Patent: No.4,
using state of charge. IEEE Trans Ind Electron 2013;60(1):191e202. 829,259A, 1989.
[11] Lin HT, Liang TJ, Chen SM. Estimation of battery state of health using proba- [42] Xiong R, Tian J, Shen WX, Sun F. A novel fractional order model for state of
bilistic neural network. IEEE Trans Ind Inf 2013;9(2):679e85. charge estimation in lithium ion batteries. IEEE Trans Veh Technol
[12] Xiong R, Li L, Tian J. Towards a smarter battery management system: a critical 2019;68(5):4130e9.
review on battery state of health monitoring methods. J Power Sources [43] Li Z, Xiong R, Mu H, He H, Wang C. A novel parameter and state-of-charge
2018;405:18e29. determining method of the lithium ion battery for electric vehicles. Appl
[13] Song JY, Wang Y, Wan C. Review of gel-type polymer electrolytes for lithium Energy 2017;207:363e71.
ion batteries. J Power Sources 1999;77(2):183e97. [44] Yao L, Wang Z. A research on the charging methods of lithium ion traction
[14] Lu L, Han X, Li J, Hua J, Ouyang M. A review on the key issues for lithium ion battery. Automot Eng 2015:72e7.
battery management in electric vehicles. J Power Sources 2013;226:272e88. [45] Xiong R, Zhang Y, Wang J, He H, Peng S, Pecht M. Lithium ion battery health
[15] Sun F, Xiong R. A novel dual-scale cell state-of-charge estimation approach for prognosis based on a real battery management system used in electric vehi-
series- connected battery pack used in electric vehicles. J Power Sources cles. IEEE Trans Veh Technol 2019;68(5):4110e21.
2015;274:582e94. [46] Engineers edge, trickle charging - battery - battery application & technology
[16] Tarascon JM, Michel A. Issues and challenges facing rechargeable lithium [EB/OL]. https://www.engineersedge.com/battery/trickle_charging.htm; 2019.
batteries. Materials for sustainable energy: a collection of peer-reviewed [47] Hussein A, Batarseh I. A review of charging algorithms for nickel and lithium
research and review articles from. Nature Publishing Group; 2011. p. 171e9. battery chargers. IEEE Trans 2011;60(3):830e8.
[17] Meena N, Baharwani V, Sharma D, Sharma A, Choudhary B, Parmar P, [48] Dong Q, Kang HX, Yan YB. Effect of charge cutoff voltage on electrochemical
Stephen RB. Charging and discharging characteristics of Lead acid and Li-ion performance of lithium ion battery. Chem Ind Eng Prog 2008;27(5):770e2.
batteries. Power Energy Syst Conf: Sustain Energy 2014:1e3. [49] Hoque M, Hannan M, Mohamed A. Optimal CC-CV charging of the lithium ion
[18] Feng X, Ouyang M, Liu X, Lu L, Xia Y, He X. Thermal runaway mechanism of battery for charge equalization controller. Adv Electrical, Electron Syst Eng Int
lithium ion battery for electric vehicles: a review. Energy Storage Mater Conf 2016.
2017;10:246e67. [50] Ayoub E, Karami N. Review on the charging techniques of a Li-Ion battery. In:
[19] Mathew M, Kong Q, Mcgrory J, Fowler M. Simulation of lithium ion battery Third international conference on technological advances in electrical, elec-
replacement in a battery pack for application in electric vehicles. J Power tronics and computer engineering (TAEECE); 2015. p. 50e5.
Sources 2017;349:94e104. [51] Liu C, Wang S, Liu Y, Tsai M. An optimum fast charging pattern search for Li-
234 Q. Lin et al. / Energy 183 (2019) 220e234

ion batteries using particle swarm optimization. In: The 6th international 2017;99. 1-1.
conference on soft computing and intelligent systems, and the 13th inter- [77] Doyle M, Fuller TF, Newman JS. Modeling of galvanostatic charge and
national symposium on advanced intelligence systems; 2012. p. 727e32. discharge of the lithium/polymer/insertion cell. J Electrochem Soc
[52] Sikha G, Ramadass P, Haran B, White R, Popov B. Comparison of the capacity 1993;140(6):1526e33.
fade of Sony US 18650 cells charged with different methods. J Power Sources [78] Thomas KE, Newman J, Darling RM. Mathematical modeling of lithium bat-
2003;122(1):67e76. teries. Netherlands: Kluwer Academic/Plenum Publishers; 2002. p. 345e92.
[53] Chen L, Roy CH, Liu C. A design of a grey-predicted Li-ion battery charge [79] Klein R, Chaturvedi N, Christensen J, Ahmed J, Findeisen R, Kojic A. State
system. IEEE Trans Ind Electron 2008;55(10):3692e701. estimation of a reduced electrochemical model of a lithium ion battery. Proc
[54] Xu Jingping. Design of constant current-constant voltage mode regulation in Am Contr Conf 2010:6618e23.
Li-ion battery charger. Semicond Technol 2011;36(4):291e5. [80] Wu H, Liu W, Chen Y, Geng H, Sun X, Yang G. Current situation of technology
[55] Liu Y, Teng J, Lin Y. Search for an optimal rapid charging pattern for lithium in charging optimization of Li-ion battery. J Power Supply 2017;15(5):
ion batteries using ant colony system algorithm. IEEE Trans Ind Electron 144e52.
2005;52(5):1328e36. [81] Asadi H, Kaboli SHA, Mohammadi A, et al. Fuzzy logic control technique in Li-
[56] Wang S, Liu Y. A PSO-based fuzzy-controlled searching for the optimal charge ion battery charger. ICEECE 2011:179e83.
pattern of Li-ion batteries. IEEE Trans Ind Electron 2015;62(5):2983e93. [82] Findeisen R, Imsland L, Allgower F, Foss B. State and output feedback
[57] Ikeya T, Sawada N, Murakami J, Kobayashi K, Hattori M, Murotani N, nonlinear model predictive control: an overview. Eur J Control 2003;9(2e3):
Yl Tomaki. Multi-step constant-current charging method for an electric 190e206.
vehicle nickel/metal hydride battery with high-energy efficiency and long [83] Methekar R, Ramadesigan V, Braatz R, Subramanian V. Optimum charging
cycle life. J Power Sources 2002;105(1):6e12. profile for lithium ion batteries to maximize energy storage and utilization.
[58] Wang H, Li G, Jiang Z, Shen J. Research on fast charge method for lithium ion ECS Trans 2010;25(35):139e46.
battery. Chin J Power Sources 2012;36(11):1616e9. [84] Suthar B, Northrop PWC, Braatz RD, Subramanian VR. Optimal charging pro-
[59] Chen Z, Xia B, Mi C, Xiong R. Loss-minimization-based charging strategy for files with minimal intercalation-induced stresses for lithium-ion batteries
lithium ion battery. IEEE Trans Ind Appl 2015;51(5):4121e9. using reformulated pseudo 2-dimensional models. J Electrochem Soc
[60] Ma J, Wang Z, Su X. Study on charging characteristics of Li-ion battery using 2014;161(11):F3144e55.
high-current pulse charging method. J Power Supply 2013;11(1):30e3. [85] Chen LR, Wu CH. Sinusoidal ripple current charging system with PLL function.
[61] Purushothaman BK, Landau U. Rapid charging of lithium ion batteries using In: IEEE 23rd International Symposium on Industrial Electronics (ISIE); 2014
pulsed currents a theoretical analysis. J Electrochem Soc 2006;153(3): 1836-1841.
A533e42. [86] Zhang Z, Wang Z, Lu X. Suppressing lithium dendrite growth via sinusoidal
[62] Chen LR. Design of duty-varied voltage pulse charger for improving Li-ion ripple current produced by triboelectric nanogenerators. Adv Energy Mater
battery-charging response. IEEE Trans Ind Electron 2009;56(2):480e7. 2019:1900487.
[63] Chen LR. A design of an optimal battery pulse charge system by frequency- [87] Lee Y, Park S. Electrochemical state-based sinusoidal ripple current charging
varied technique. IEEE Trans Ind Electron 2007;54(1):398e405. control. IEEE Trans Power Electron 2015;30(8):4232e43.
[64] Xiong R, Yu Q, Shen WX, Lin C, Sun F. A sensor fault diagnosis method for a [88] Bessman A, Vadivelu S, Wallmark O, Soares R, Svens P, Ekstrom H,
lithium-ion battery pack in electric vehicles. IEEE Trans Power Electron 2019. Lindbergh G. Challenging sinusoidal ripple-current charging of lithium-ion
1-1. batteries. IEEE Trans Ind Electron 2018;65(6):4750e7.
[65] Zhang Y, Liu H, Lu M. Modeling of unified charging of lithium ion batteries. [89] Vadivelu S. Investigation of sinusoidal ripple current charging techniques for
Chin J Power Sources 2015;39(2):283e4. Li-ion cells. 2016. p. 138.
[66] Zhang SS. The effect of the charging method on the cycle life of a Li-ion [90] Cho SY, Lee IO, Baek JI, Moon GW. Battery impedance analysis considering DC
battery. J Power Sources 2006;161(2):1385e91. component in sinusoidal ripple-current charging. IEEE Trans Ind Electron
[67] Gering K, Sazhin S, Jamison D, Michelbacher C, Liaw B, Dubarry M, Cugnet Ml. 2015;63(3):1561e73.
Investigation of path dependence in commercial lithium ion cells chosen for [91] Xu J, Wang J, Li S, Cao B. A method to simultaneously detect the current sensor
plug-in hybrid vehicle duty cycle methods. J Power Sources 2011;196(7): fault and estimate the state of energy for batteries in electric vehicles. Sensors
3395e403. 2016;v16. 1328.
[68] Notten PHL, het Veld JHGO, van Beek JRG. Boost charging Li-ion batteries: a [92] Chaoui H. State of charge and state of health estimation for lithium batteries
challenging new charging concept. J Power Sources 2005;145(1):89e94. using recurrent neural networks. IEEE Trans Veh Technol 2017;99. 1-1.
[69] Savoye F, Venet P, Millet M, Groot J. Impact of periodic current pulses on Li- [93] Xiong R, Li L, Li Z, Yu Q, Mu H. An electrochemical model based degradation
ion battery performance. IEEE Trans Ind Electron 2012;59(9):3481e8. state identification method of Lithium-ion battery for all-climate electric ve-
[70] Rahimian SK, Rayman SC, White RE. Maximizing the life of a lithium ion cell hicles application. Appl Energy 2018;219:264e75.
by optimization of charging rates. J Electrochem Soc 2010;157(12):A1302e8. [94] Xiong R, Zhang Y, He H, Zhou X, Pecht M. A double-scale, particle-filtering,
[71] Wang X, Xu J, Zhao Y. Wavelet based denoising for the estimation of the state energy state prediction algorithm for lithium-ion batteries. IEEE Trans Ind
of charge for lithium-ion batteries. Energies 2018;11:1144. 05-04 2018. Electron 2018;65(2):1526e38.
[72] Ye M, Gong H, Xiong R, Mu H. Research on the battery charging strategy with [95] Yang F, Qiao Y, Gan D, Wang Q, Chen W. Lithium ion battery polarization
charging and temperature rising control awareness. IEEE Access 2018;6: characteristics at different charging methods. Trans China Electrotech Soc
64193e201. 2017;32(12):171e8.
[73] Guo S, Xiong R, Wang K, Sun F. A novel echelon internal heating strategy of [96] Jiang J, Liu Q, Zhang C, Zhang W. Evaluation of acceptable charging current of
cold batteries for all-climate electric vehicles application. Appl Energy power Li-ion batteries based on polarization characteristics. IEEE Trans Ind
2018;219:256e63. Electron 2014;61(12):6844e51.
[74] Inoa E, Wang J. PHEV charging strategies for maximized energy saving. IEEE [97] Jiang J, Zhang C, Wen J, Zhang W, Sharkh S. An optimal charging method for
Trans Veh Technol 2011;60(7):2978e86. Li-ion batteries using a fuzzy-control approach based on polarization prop-
[75] Perez HE, Hu X, Dey S, Moura SJ. Optimal charging of Li-ion batteries with erties. IEEE Trans Veh Technol 2013;62(7):3000e9.
coupled electro-thermal-aging dynamics. IEEE Trans Veh Technol 2017;66(9): [98] Xiong R, Chen H, Wang C, Sun F. Towards a smarter hybrid energy storage
7761e70. system based on battery and ultracapacitor - a critical review on topology and
[76] Zou C, Hu X, Wei Z, Wik T, Egardt B. Electrochemical estimation and control energy management. J Clean Prod 2018;202:1228e40.
for lithium ion battery health-aware fast charging. IEEE Trans Ind Electron