energies

Article
A Cubature Particle Filter Algorithm to Estimate the
State of the Charge of Lithium-Ion Batteries Based on
a Second-Order Equivalent Circuit Model
Bizhong Xia, Zhen Sun, Ruifeng Zhang * and Zizhou Lao
Division of Advanced Manufacturing, Graduate School at Shenzhen, Tsinghua University, Shenzhen 518055,
China; xiabz@sz.tsinghua.edu.cn (B.X.); sz15@mails.tsinghua.edu.cn (Z.S.); lzz15@mails.tsinghua.edu.cn (Z.L.)
* Correspondence: zrf223@126.com; Tel.: +86-188-9836-6838

Academic Editor: Izumi Taniguchi

Received: 13 January 2017; Accepted: 23 March 2017; Published: 1 April 2017

Abstract: The state of charge (SOC) is the residual capacity of a battery. The SOC value indicates
the mileage endurance, and an accurate SOC value is required to ensure the safe use of the battery
to prevent over- and over-discharging. However, unlike size and weight, battery power is not
easily determined. As a consequence, we can only estimate the SOC value based on the external
characteristics of the battery. In this paper, a cubature particle filter (CPF) based on the cubature
Kalman filter (CKF) and the particle filter (PF) is presented for accurate and reliable SOC estimation.
The CPF algorithm combines the CKF and PF algorithms to generate a suggested density function
for the PF algorithm based on the CKF. The second-order resistor-capacitor (RC) equivalent circuit
model was used to approximate the dynamic performance of the battery, and the model parameters
were identified by fitting. A dynamic stress test (DST) was used to separately estimate the accuracy
and robustness of the CKF and the CPF algorithms. The experimental results show that the CPF
algorithm exhibited better accuracy and robustness than the CKF algorithm.

Keywords: state of charge; cubature Kalman filter; cubature particle filter; electric vehicles;
lithium-ion battery

1. Introduction
With the problems due to industrial development and environmental pollution, electric vehicles
are becoming more popular as an environmentally-friendly mode of transportation. The battery
management system (BMS) is an important component of electric vehicles. The main role of the BMS
is to provide a reasonable algorithm to utilize the function of the battery at an appropriate level for the
working environment and provide feedback to the driver [1,2]. The state of charge (SOC) of a battery
is functionally equivalent to a fuel gauge indicating the life of the battery pack in a battery-containing
electric vehicle. SOC estimation is an important function of the BMS, and is defined as the ratio of
the remaining capacity of the battery to the rated capacity. Accurate SOC estimation can maximize
the performance of the battery and protect the battery to prevent overcharge and over discharge.
The SOC value should provide drivers information about the status of the electric vehicle so that
drivers can develop a reasonable driving plan. However, it is difficult to measure SOC directly and it is
typically estimated from direct measurement variables. Some approaches have been tested and found
to provide precise estimation of battery SOC, but these methods are protracted, costly, and interrupt
main battery performance. It is impossible to make intuitive SOC value measurements. Although SOC
value exhibits a monotonous relationship with the battery open circuit voltage, the SOC value is very
sensitive to the change of battery voltage, and even small voltage changes will translate to significant

Energies 2017, 10, 457; doi:10.3390/en10040457 www.mdpi.com/journal/energies

Energies 2017, 10, 457 2 of 15

changes in the SOC value. Overall, it is a significant challenge to obtain an accurate value of SOC.
For this reason, estimation of the SOC value is a preferred approach.
Previous studies have proposed many methods for SOC estimation, such as the Coulomb counting
method (ampere-hour (Ah) integration method) [3–5], the open circuit voltage method [5,6], the BP
(back-prorogation) neural network algorithm [7], Kalman filtering algorithm [8,9], and the particle
filter (PF) method [10]. Additionally, some improved algorithms based on these methods have been
proposed, such as the adaptive Kalman filter [11], the unscented Kalman filter (UKF) [12–15], the strong
tracking cubature Kalman filter (STCKF) [16], and the unscented PF [17]. The Coulomb counting
method is widely used for SOC estimation of electric vehicles because it is a simple algorithm. This
method uses the SOC formula to compute the current integration and obtain the charged or discharged
capacity. However, the current sensor itself has an intrinsic measurement error. During estimation,
this current measurement error is integrated, and the error increases. The open-circuit voltage method
is based on the monotonous relationship of the open-circuit voltage between the battery and the SOC.
In this method, the open-circuit voltage must be measured for a sufficient period of time. An additional
challenge of accurate measurement is that the SOC value for a lithium-ion battery is highly sensitive to
changes in open circuit voltage. Rumelhart and McCelland proposed a neural network algorithm using
an error BP training algorithm [18]. The BP network is the simplest of the human brain abstraction
and simulation models using input and output samples and a corresponding training network to
achieve an input-output mapping function relationship. However, the neural network method requires
a significant amount of sample data for training, and the estimation error is greatly affected by the
quality of the training data and the training method. Additionally, the hardware requirements are high.
The Kalman filter (KF) is an autoregressive optimal data processing algorithm proposed by
Kalman in 1960 [19]. Its core idea is to make the best estimate of the minimum variance in the
system state. The KF algorithm overcomes the error accumulation effect of the coulomb counting
method that occurs with increased time. The KF algorithm does not depend on an accurate initial
SOC value, but can improve the SOC value accuracy. However, the accuracy of this method depends
on the establishment of a battery equivalent model, and some physical properties of the battery
model are nonlinear. The EKF algorithm [20,21] and the UKF algorithm are improved KF algorithms.
The EKF algorithm implements recursive filtering by linearizing nonlinear functions [22], and the
UKF algorithm applies nonlinear system equations to the standard Kalman filter system by means of
unscented transformation (UT). UT is a mathematical function used to estimate the result of applying
a given nonlinear transformation to a probability distribution that is characterized by a finite set of
statistics. Compared with the EKF algorithm, the UKF algorithm exhibits higher accuracy and has a
wider application range, making it well-suited for solving nonlinear problems [23].
The PF method is a random sampling-based filtering method used to solve non-linear
non-Gaussian problems [24,25]. The rationale of this method is to use a series of weighted random
sample sets (particles) in the state space to approximate the posterior probability density function
of the system states. Although the PF algorithm is relatively simple, there may be significant loss of
accuracy during sampling. The main way to decrease the degradation of the PF method is to increase
the number of particles and resample. However, re-sampling will reduce the diversity of particles, and
a large increase in the number of particles will greatly increase the required time of the calculation.
Therefore, it is necessary to select the importance density function of the PF algorithm.
The cubature Kalman filter (CKF) algorithm is a kind of filtering algorithm that is free of
differentiation and high in filtering precision [16,26–28]. In this paper, we present the cubature
particle filter (CPF) method for SOC estimation. The CKF algorithm directly calculates the mean
and variance of the state using the numerical integration method based on the cubature principle,
and generates the importance density function of the PF algorithm using the mean and the variance.
This approach combines aspects of the CKF algorithm and the PF algorithm. Through the dynamic
stress test (DST) and New European Driving Cycle (NEDC) tests, we verified that the accuracy and
robustness of the CPF algorithm is superior to that of the CKF algorithm.
Energies 2017, 10, 457 3 of 15

The remainder of this paper is arranged as follows. The second-order resistor-capacitor (RC)
equivalent circuit model and the model parameters are presented in Section 2; Section 3 introduces the
CKF algorithm and the CPF algorithm in detail; Section 4 describes the schematic of the battery test;
And in Section 5, the verification results and comparisons of the CKF algorithm and CPF algorithm for
accuracy and robustness are presented, and as well as the summary and conclusions.
Energies 2017, 10, 457 3 of 15

2. Battery Equivalent Circuit Model and Identification of Model Parameters

test; And in Section 5, the verification results and comparisons of the CKF algorithm and CPF
algorithm for accuracy and robustness are presented, and Section 6 presents the summary and
2.1. Battery Equivalent Circuit Model
conclusions.
Modeling of battery performance SOC is often required during SOC estimation. Many models
2. Battery Equivalent Circuit Model and Identification of Model Parameters
have been proposed, including the electrochemical model [29], electrochemical-thermal coupling
model [30,31], andEquivalent
2.1. Battery performanceCircuit models
Model that are often used to describe the external characteristics of
the battery during
Modeling of battery performance SOC iscircuit
operation. The equivalent model consists
often required during SOC of estimation.
resistors, capacitors,
Many modelsvoltage
sources,have
and been
otherproposed,
circuit components
including the electrochemical model [29], electrochemical-thermal battery.
to simulate the dynamic characteristics of the couplingFor the
use of circuit
modeland mathematical
[30,31], and performance methods
modelsfor that analysis,
are often the
usedequivalent circuit
to describe the model
external is the most
characteristics of widely
the battery
used battery model. during operation. The equivalent circuit model consists of resistors, capacitors, voltage
Thesources,
accuracy andand
othercomputational
circuit components to simulate
complexity ofthe dynamic characteristics
estimating SOC vary forofdifferent
the battery. For themodels.
battery
use of circuit and mathematical methods for analysis, the equivalent circuit model is the most widely
The selection of circuit model should be based on whether the equivalent circuit model can reasonably
used battery model.
reflect the characteristics
The accuracy and of the battery andcomplexity
computational the CPU of processing
estimatingpower
SOC vary of thefor BMS. A more
different batterycomplex
circuit model
models.may be able to
The selection accurately
of circuit modelrepresent
should be based the performance
on whether thecharacteristics
equivalent circuitofmodel
the battery
can but
use of a reasonably
more complex reflect model can increase
the characteristics of thethe computational
battery complexity
and the CPU processing powerof the SOC
of the BMS. calculation
A more and
increasecomplex circuit model
the workload of themay CPU.be able to accurately
However, represent the
a simplified performance
equivalent characteristics
circuit model may of thenot well
battery but use of a more complex model can increase the computational complexity of the SOC
represent the performance characteristics of the battery, lowering the SOC estimation accuracy.
calculation and increase the workload of the CPU. However, a simplified equivalent circuit model
Themay
selection of a reasonable equivalent circuit model is important for SOC estimation. Many
not well represent the performance characteristics of the battery, lowering the SOC estimation
differentaccuracy.
equivalent circuit models have been proposed, such as the resistance model [32,33],
the first-orderTheRC equivalent
selection circuit model
of a reasonable equivalent[6,33], the model
circuit second-order RCfor
is important equivalent circuitMany
SOC estimation. model [34],
and otherdifferent
models equivalent
[35–37].circuit
The models
resistorhavemodel beenis proposed,
simple suchand as the resistance model
computationally [32,33],
small, theitfirst-
thus requires a
order RC equivalent circuit model [6,33], the second-order RC equivalent
low CPU load but provides poor accuracy. The complex equivalent circuit models exhibit high precision circuit model [34], and other
models [35–37]. The resistor model is simple and computationally small, thus it requires a low CPU
for determining the dynamic voltage characteristic of a battery, but the model is computational complex
load but provides poor accuracy. The complex equivalent circuit models exhibit high precision for
and thedetermining
CPU load the is heavy.
dynamicThe first-order
voltage characteristic RC of equivalent
a battery, butcircuit model
the model and the second-order
is computational complex RC
equivalent
and circuit
the CPUmodelload isachieve
heavy. The a good balance
first-order RC between
equivalent computational
circuit model andcomplexity and the
the second-order RC ability
to provide approximate
equivalent accuracy
circuit model achieveof abattery characteristics,
good balance with the second-order
between computational complexity and equivalent
the ability circuit
exhibitedto provide approximate
better accuracy foraccuracy of batteryof
the estimation characteristics, with the second-order
battery characteristics. equivalent
Therefore, circuit
for this study, we
exhibited better accuracy for the estimation of battery characteristics. Therefore, for this study, we
selected a second-order RC equivalent circuit model. The second-order RC equivalent circuit model
selected a second-order RC equivalent circuit model. The second-order RC equivalent circuit model
includes a voltage source Uoc (SOC) with monotonicity to the SOC and a resistor R0 and two sets of
includes a voltage source Uoc(SOC) with monotonicity to the SOC and a resistor R0 and two sets of
resistorsresistors
and capacitors (( R1 , ((
and capacitors CR1 ),C;) ; ((R
1 1
, C))
R22,C 2
2 ))ininparallel
parallel from
from the the circuit.
circuit. The structure
The structure of the circuit
of the circuit
model ismodel
shown in Figure 1.
is shown in Figure 1.

Figure 1. Model structure of a lithium-ion battery.

Figure 1. Model structure of a lithium-ion battery.
Energies 2017, 10, 457 4 of 15

SOC is usually defined as shown in Equation (1):

Rt
t0 idt • i (t)
SOC (t) = SOC (t0 ) − ⇒ SOC (t) = − (1)
QN QN

where SOC(t) and SOC(t0 ) represent SOC values at times t and t0 , respectively; i(t) represents the value
•
of the current at time k; and QN indicates the rated capacity of the battery. SOC (t) is the derivative
of SOC(t).
According to Kirchhoff's law, the following equations (Equations (2)–(4)) are obtained from the
second-order RC equivalent circuit model:

U1 dU • i (t) U1
+ C1 1 = i (t) ⇒ U1 = − (2)
R1 dt C1 C1 R1

U2 dU • i (t) U2
+ C2 2 = i (t) ⇒ U2 = − (3)
R2 dt C2 C2 R2
Ut = Uoc (SOC ) − U1 − U2 − R0 i (t) (4)
• •
U1 and U2 denote the terminal voltage of C1 and C2 , respectively; U1 and U2
are the derivatives of
U1 and U2 , respectively; Ut and i(t) represent the value of the terminal voltage and current, respectively.
Uoc (SOC) indicates the open circuit voltage of the battery (under the same environmental conditions,
the open-circuit voltage value and the SOC value are monotonous.)
The equation of state for the discretization of the system is:

− QTN
      
SOC (k) 1 0 0 SOC (k − 1)
  0 1− T 0   T 
 U1 (k)  =   U1 (k − 1)  +  C1 i (k) (5)
 
C1 R1
T T
U2 (k) 0 0 1− C2 R2 U2 (k − 1) C 2

Taking the terminal voltage as the observed value, we obtain the observed equation:

Ut (k ) = Uoc (SOC ) − R0 i (k) − U1 (k) − U2 (k ) (6)

2.2. Parameter Identification

The second-order system model of the battery was used as follows. For SOC estimation, some
parameters in the model must be identified in advance, including Uoc (SOC) and the values of
{ R0 , R1 , R2 , C1 , C2 }. Under the same temperature conditions, an ICR18650-22F lithium-ion battery
was tested as follows: (1) the battery was charged to the upper limit cut-off voltage at 4.2 V, and left
standing for 2 h; (2) the battery was discharged to the lower limit cutoff voltage at 2.75 V, and the
discharged capacity was recorded as the battery capacity QN ; (3) repeat step (1), and the battery was
continuously discharged at a rate of 0.1C, and was allowed to stand for 2 h after discharging the battery
at a rated capacity of 10%; and (4) repeat step (3) until the battery is completely discharged.
As shown in Figure 2, the blue-solid triangles reflect the relationship between Uoc (SOC) and
SOC after 2 h of standing in the test step. The experimental results were fitted by a sixth-order
polynomial fitting curve, and the results were approximate to the relationship between Uoc (SOC) and
SOC. The values of { R0 , R1 , R2 , C1 , C2 } were obtained by fitting the data for a certain discharge and the
standing process of the battery in the experimental step by exponential function fitting. The identified
parameters are listed as follows:
R0 = 0.0361Ω, R1 = 0.0218Ω, R2 = 0.0051Ω, C1 = 1658.8F, C2 = 87751F.
Energies 2017, 10, 457 5 of 15
Energies 2017, 10, 457 5 of 15

Figure 2. Uoc(SOC) identification. SOC: state of charge.

Figure 2. Uoc (SOC) identification. SOC: state of charge.
3. Flow of State of Charge Estimation Algorithm
3. Flow of State of Charge Estimation Algorithm
3.1. Cubature Kalman Filtering Algorithm (CKF)
3.1. Cubature Kalman Filtering Algorithm (CKF)
Consider the following nonlinear non-Gaussian discrete-time dynamic systems:
Consider the following nonlinear non-Gaussian
 x
discrete-time dynamic systems:
= f (x ) + w k +1 k k
 (7)
(  y k +1 = h ( x k +1 ) + v k +1
x k +1 = f ( x k ) + w k
where xk +1 and yk +1 are the system state vector and the measured value at time k + 1; f (⋅) and (7)
y k +1 = h ( x k +1 ) + v k +1
h (⋅) are the system state transition model function and the measurement model function,
respectively; wk and vk +1 are process noise and observation noise, respectively.
where xk+1 and yk+1 are the system state vector and the measured value at time k + 1; f (·) and h(·) are
the system
(1) state transition model function and the measurement model function, respectively; wk and
Initialization:
vk+1 are process noise and observation noise, respectively.
xˆ0 = E[ x0 ], P0 = E[( x0 − xˆ0 )( x0 − xˆ0 ) T ] (8)
(1) Initialization:
(2) Time update: x̂0 = E[ x0 ], P0 = E[( x0 − x̂0 )( x0 − x̂0 )T ] (8)
Calculate the cubature points:
(2) Time update:
Pk = S k S kT (9)
Calculate the cubature points:
X j , k = S k ξ i + xˆ k i = 1, 2, , 2 n (10)
Pk = Sk SkT (9)
where n represents the state dimension, the total number of volume points is 2 times the state
dimension, and the volume X j,k =is S
point k ξ i + x̂k as shown
calculated · · · , 2n (11):
i = 1,in2,Equation (10)

where n represents the state dimension,

ξ j = nthe
[1] jtotal
j = 1,number
,2n of volume points is 2 times
(11)the state
dimension, and the volume point is calculated as shown in Equation (11):
where [1] j indicates that the point is centered at the j-th point of [1] , and the symbol [1] is a
√
complete set of all symmetric points,
ξ j =denoting
n [1] j the set
j =of1,points generated by the full permutation
· · · , 2n (11)
of the elements of the n-dimensional unit vector e = [1, 0,  , 0]T and the change of the element
wheresymbol.
[1] j indicates that the point is centered at the j-th point of [1], and the symbol [1] is a
complete set ofthe
Calculate allcubature points
symmetric that are
points, propagated
denoting the through the state
set of points equation:by the full permutation
generated
of the elements of the n-dimensionalX unit f ( x j , k ) e = [1, 0, · · · , 0]T and the change
j , k +1| k =vector (12) of the
element symbol.
Calculate the square root of state prediction and variance prediction:
Calculate the cubature points that are propagated
1 2 n ithrough the state equation:
xˆ k +1| k =
2n
x
i =1
k + 1| k (13)
X j,k+1|k = f ( x j,k ) (12)

Calculate the square root of state prediction and variance prediction:

1 2n i
2n i∑
x̂k+1|k = x k +1| k (13)
=1
Energies 2017, 10, 457 6 of 15

1 2n i T
2n i∑
Pk+1|k = Xk+1|k ( Xki +1|k ) − x̂k+1|k ( x̂k+1|k )T + Qk (14)
=1

where Qk denotes the process noise covariance at time k.

(3) Measurement update:

Calculate the cubature points:
Pk+1|k = Sk+1|k SkT+1|k (15)

x j,k+1|k = Sk+1|k ζ j + x̂k+1|k (16)

Calculate the volume points propagated through the measurement equation and prediction:

y j,k+1 = h( x j,k+1|k ) (17)

1 2n
2n i∑
ŷk+1 = y j,k+1 (18)
=1

Calculate the predicted covariance:

1 2n
2n i∑
Py,k+1 = y j,k+1 (y j,k+1 )T − ŷk+1 (ŷk+1 )T + Rk+1 (19)
=1

1 2n
2n i∑
Pxy,k+1 = x j,k+1|k ( x j,k+1|k )T − x̂ j,k+1|k ( x̂ j,k+1|k )T (20)
=1

where Rk+1 denotes the measurement noise covariance at time k + 1.

(4) Calculate the gain matrix, the state of time k + 1 and the square root of the variance estimates:

Kk+1 = Pxy,k+1 ( Py,k+1 )−1 (21)

x̂k+1 = x̂k+1|k + Kk+1 (yk+1 − ŷk+1 ) (22)

Pk+1 = Pk+1|k − Kk+1 Py,k+1 KkT+1 (23)

3.2. Cubature Particle Filtering Algorithm

CPF applies the CKF to the PF framework. At time k, the mean value and variance of
the importance density function are calculated by the CKF algorithm based on the observation
data [16,26]. A new posterior probability density distribution is generated, and a new posterior
probability distribution can re-generate the new particles, calculate the weights of the particles, perform
normalization, and then perform resampling to complete the state estimation. The implementation of
the CPF algorithm is as follows:
(i ) (i )
(1) Initialize to obtain initial particles x̂0 and corresponding variance P̂0 , i.e., to extract particles
(i )
x0 (1, · · · , N ) from the initial state probability distribution density p( x0 ), and:

(i ) (i )
x̂0 = E[ x0 ] (24)

(i ) (i ) (i ) (i ) (i ) T
P̂0 = E[( x0 − x0 )( x0 − x0 ) ] (25)

(2) When k = 1, 2, · · · , proceed as follows:

Energies 2017, 10, 457 7 of 15

(i ) (i )
If the particle set with equal weights (1/N) at time k − 1 is { xk−1 , P̂k−1 , i = 1, · · · , N }, the particles
(i ) (i )
are subjected to cubature Kalman filtering, and the particle set and covariance ( x̂k , P̂k ) are obtained.
(i ) (i )
The density function of choice is the Gaussian distribution function with x̂k as mean and P̂k as
variance as follows: 
(i )  (i ) (i ) (i )
q( xk xk−1 , z1:k ) = N ( x̂k , P̂k ) (26)

The importance density function defined in the formula is used to generate the predicted particle
(i ) (i )
set and the corresponding variance { x̂k , P̂k , i = 1, · · · , N }. Because the sampling points of each
particle have different Gaussian distributions and any non-Gaussian distribution can include a series
of different Gaussian distributions, it is reasonable to assume the importance density function as a
Gaussian distribution.
The calculation of the weight of each particle:
 
 (i ) ( i )  ( i −1)
(i ) ( i −1)
p(zk xk ) p( xk xk )
wk = wk 
( i )  ( i −1)
(27)
q( xk xk , zk )

Normalized weights:
(i )
(i ) wk
wk = (28)
N ( j)
( ∑ wk )
j =1

Resampling is performed as described in [38].

Resampling judgment condition:

1
N̂e f f = < Nthreshold (29)
N 2
∑ (wik )
i =1

(i ) (i )
The particle set { x̂k , P̂k , i = 1, · · · , N } is resampled corresponding to the normalized weight
(i ) (i ) (i )
wk to obtain the set of particles { xk , P̂k , i = 1, · · · , N } with equal weight 1/N.
Output the state and variance estimates at time k:

1 N (i )
N i∑
x̂k = xk (30)
=1

1 N ( i ) ( i )T
N i∑
P̂k = xk xk − x̂k x̂kT (31)
=1

The cubature particle filtering process outlined in Figure 3.

4. Experimental Configurations and Analysis

4.1. Experimental Configurations

The schematic of the battery test bench is shown in Figure 4. It consists of: (1) lithium-ion
batteries; (2) a control board (Sunwoda, Shenzhen, China) for the control of battery charge/discharge;
(3) a personal computer with monitoring software for data sampling and MATLAB R2010a (MathWorks,
Natick, MA, USA)) for data analysis; (4) a DC contactor (Schneider, Beijing, China) for charge/discharge
switching; (5) a power supply (ITECH IT6952A, ITECH Electronics, Nanjing, China) for cell charging
and (6) a programmable electric load (ITECH IT8510, ITECH Electronics, Nanjing, China) for
cell discharging.
Energies 2017, 10, 457 8 of 15
Energies 2017, 10, 457 8 of 15
Energies 2017, 10, 457 8 of 15

xˆ0( i ) = E[ x0( i ) ] P0(i ) = E[( x0(i ) − xˆ0(i ) )( x0(i ) − xˆ0(i ) )T ]

xˆ0( i ) = E[ x0( i ) ] P0(i ) = E[( x0(i ) − xˆ0(i ) )( x0(i ) − xˆ0(i ) )T ]

Pk(i ) = S k( i ) ( Sk(i ) )T X (ji,k) = Sk( i )ξi + xˆk( i ) i = 1, 2, , 2n

(i )
P = S (S ) (i ) (i ) T
X (ji,k) = S( i )k( i )ξi + 1xˆk( i )2 n i =( i1, 2, , 2n X k( i+)1 = X kj+(1i )
X k (i )
= f (X )
j , k +1|k
k k (i )
j ,k
xˆk +1|k =
2 n
 j2=n1
X j ,k) +1|k
XWk( i+i)1 ==X1kj+(1i )
1
X (j i,k) +1|k =1 f 2(nX (j i,k) ) xˆk( i+)1|k =  X (ji,k) +1|k k +1
N1
Pk +1|k =  X ki +1|k ( X ki +1|k )T − xˆk +1|k2(nxˆk +j 1|=1k )T + Qk
21n i2=n1 i
Wki+1 =
N
Pk +1|k =  X k +1|k ( X ki +1|k )T − xˆk +1|k ( xˆk +1|k )T + Qk
2n i =1

Pk +1|k = Sk +1|k SkT+1|k x j ,k +1|k = Sk +1|k ζ j + xˆk +1|k

Pk +1|k = Sk +1|k SkT+1|k x j ,k +1|k = S1k +1|2k ζn j + xˆk +1|k
y j ,k +1 = h( x j ,k +1|k ) yˆ k +1 =  y j ,k +1
21n i2=n1 X k(i+)1 ~ q i = N ( xˆki +1 , pˆ ki +1 )
y j ,k +1 = h( x j ,k +1|k )
1 2n
ˆ
y k +1 =  y j ,k +1
2n i =1T X k(i+)1 ~ q i = N ( xˆki +1 , pˆW
i
 j ,k +1 j ,k +1 k +1 )
T i
Py ,k +1 = y ( y ) − ˆ
y ( ˆ
y k +1 ) + Rk +1
21n i2=n1
k +1 Wki+1 = q iWki+1 Wki+1 = N k +1
Py ,k +1 =  j ,k +1 j ,k +1 T k +1 k +1 )T + Rk +T1
21n i2=n1
y ( y ) T
− ˆ
y ( ˆ
y
W i = q iW i W i = W
Wki+1n
k +1
 x j ,k +1|k ( x j ,k +1|k ) − xˆ j ,k +1|k ( xˆ j ,k +1|k )
k +1 k +1 k +1
Pxy,k +1 = nN
=1
21n i2=n1 W n
k +1
Pxy,k +1 = 1 x j ,k +1|k ( x j ,k +1|k )T − xˆ−1j ,k +1|k ( xˆ j ,k +1|k )T
2n i =K
n =1

=Pk +1 (P )
xy, k +1 y , k +1

K k +1 = Pxy, k +1 ( Py ,k +1 )−1

xˆki +1 = xˆki +1|k + K k +1 ( yk +1 − yˆ ki +1 )

i
ˆki +1|i k + K k +1 ( yk +1 − yˆTki +1 )
kˆ+1i ==xP
xˆP
k +1 k +1|k − K k +1 Py , k +1 K k +1

Pˆ i = P i − K P K T
k +1 k +1|k k +1 y , k +1 k +1

Figure 3. Cubature particle filtering (CPF) algorithm process.

Figure 3. Cubature particle filtering (CPF) algorithm process.
Figure 3. Cubature particle filtering (CPF) algorithm process.

Figure 4. Schematic of the battery test bench.

Figure 4. Schematic of the battery test bench.
Figure 4. Schematic of the battery test bench.

Samsung ICR18650-22F lithium-ion batteries (Samsung SDI, Seoul, Korea) were used, with a
nominal voltage and nominal capacity of 3.62 V and 2.2 Ah, respectively.
 Energies
Energies 2017, 10,2017,
45710, 457 9 of 15 9 of 15

EnergiesSamsung
2017, 10, 457ICR18650-22F
lithium-ion batteries (Samsung SDI, Seoul, Korea) were used, with
9 ofa15
nominal voltage
4.2. Experimental Analysisand nominal capacity of 3.62 V and 2.2 Ah, respectively.
Samsung ICR18650-22F lithium-ion batteries (Samsung SDI, Seoul, Korea) were used, with a
4.2. Experimental
nominal voltage Analysis
4.2.1. Dynamic Stressand nominal capacity of 3.62 V and 2.2 Ah, respectively.
Test
The4.2.1.
4.2. DSTDynamic
Experimental
tests theStress Test
Analysis
function of a battery under dynamic test conditions that are generated from
the simplification
The DSTof federal
tests urban of
the function driving stress
a battery under(FUDS)
dynamictest [39].
test To testthat
conditions whether the CPFfrom
are generated algorithm
4.2.1. simplification
Dynamic Stress Test urban driving stress (FUDS) test [39]. To test whether the CPF algorithm
is moretheaccurate for SOC of federal
estimation than the CKF algorithm, we used the DST to conduct battery
is more
charge/discharge accurate
The DST testsfor
tests. SOC
theThe estimation
function
resultsof are than
a battery
shownthe CKF
under algorithm,
dynamic
in Figure Inwe
5. test used the DST
conditions
the that to
experiment, conduct
are12,000 sbattery
generatedof from
data were
charge/discharge
the simplification tests.
of Theurban
federal resultsdriving
are shown in(FUDS)
stress Figure 5. In[39].
test the experiment,
To test 12,000
whether thes CPF
of data were
algorithm
collected for analysis as shown in Figure 5a. A representative experimental cyclic unit is shown in
Figureiscollected for analysis
more accurate
5b. for SOCas shown in Figure
estimation than 5a.
theACKFrepresentative
algorithm, experimental
we used the cyclic
DST tounit is shown
conduct in
battery
Figure 5b.
charge/discharge tests. The results are shown in Figure 5. In the experiment, 12,000 s of data were
collected for analysis as shown in Figure 5a. A representative experimental cyclic unit is shown in
Figure 5b.

(a) (b)

Figure 5. Dynamic stress testing (DST): (a) current profile and (b) magnified plot of the current profile.
Figure 5. Dynamic stress testing (DST): (a) current profile and (b) magnified plot of the current profile.
(a) (b)
Figure 6 shows the performance of the algorithm under the DST. The red solid-line and green
Figure 5. Dynamic stress testing (DST):
solid-line
Figure 6 shows are thetheestimated results
performance of (a)
from the
the current profile
CPF and
algorithm EKFandmethod,
under(b) magnified
DST.plot
therespectively.
Theof the
The
red current profile.
capacity
solid-line of and
the green
battery was obtained experimentally as described in Section 2.2. According to Equation (1), the
solid-lineFigure
are the estimated
6 shows
results from thealgorithm
CPF andunder EKF the method, respectively. The capacity of
reference SOC valuethe
shownperformance
in Figure 6a ofwas
the obtained by integrating DST. The red
the charge andsolid-line
dischargeand green
current
the battery
solid-linewas
per second areinobtained
thetheestimated
experimentally
DST. To more results asCPF
fromcompare
easily the described
and
the EKFin
methods,
Section
method, 2.2. According
respectively.
an initial SOC valueThe
to Equation
of 1 capacity
was usedoffor the (1),
the reference
battery SOC
was value
obtained shown in
experimentally Figure as 6a was
described obtained
in Sectionby integrating
2.2. According
the CKF and CPF algorithms, equal to the initial measured value. The experimental results are shown theto charge
Equation and(1),discharge
the
current inper
referencesecond
Figure SOC
6b and in the
value DST.
shown
Table ToFigure
in
1. The more 6a
average easily compare
was obtained
absolute the
by
error and methods,error
integrating
maximum ancharge
the initial SOC
of the and
CPF value ofcurrent
discharge
algorithm 1 was used
were
per second
0.5% and in the
2.6%, DST.
and theTo more
average easily
absolutecompare
error the
and methods,
maximum an initial
error
for the CKF and CPF algorithms, equal to the initial measured value. The experimental results of SOC
the CKFvalue of
algorithm1 was used
were 1%for are
the
andCKF and
3.8%. CPF
Thus, algorithms,
the CPF equal
algorithm to
is the
more initial measured
accurate than value.
the CKF The experimental
algorithm.
shown in Figure 6b and Table 1. The average absolute error and maximum error of the CPF algorithm As shownresults
in are
Figureshown
7a,
inthe
were 0.5% blue6b
Figure solid-line
and 2.6%, and and represents
Table 1. The
the averagethe measured
average absolute
absolute voltage
error
error andmaximum
and
and the red solid-line
maximum error of
error represents
of the
theCPF the voltage
CKFalgorithm
algorithm werewere 1%
estimated
0.5% and 2.6%,by theandCPFthemethod.
average The blue solid-line
absolute error andand the red error
maximum solid-line exhibit
of the the same trend
CKF algorithm were of 1%
and 3.8%. Thus,
change,
the CPF algorithm is more accurate than the CKF algorithm. As shown in Figure 7a,
and 3.8%.with
Thus, less
thethan
CPF0.04 voltage is
algorithm error,
more asaccurate
shown inthan
Figure
the 7b.
CKF algorithm. As shown in Figure 7a,
the blue solid-line represents the measured voltage and the red solid-line
the blue solid-line represents the measured voltage and the red solid-line represents the voltage
represents the voltage
estimated by the
estimated CPFCPF
by the method.
method.The Theblueblue solid-line andthe
solid-line and thered
red solid-line
solid-line exhibit
exhibit the same
the same trendtrend
of of
change, withwith
change, less less
thanthan 0.040.04
voltage
voltageerror,
error,asasshown
shown in Figure7b.
in Figure 7b.

Figure 6. SOC estimation as determined by DST: (a) SOC estimation; and (b) SOC estimation error.
Energies 10, 457
2017,2017,
Energies 10, 457 10 of10
15of 15

EnergiesFigure
2017, 10,
6. 457
SOC estimation as determined by DST: (a) SOC estimation; and (b) SOC estimation error.10 of 15
Table 1. Comparison of the performance of the cubature Kalman filtering (CKF) and CPF algorithms in
the DST
Tabletest.
Figure1.6.Comparison
SOC estimation
of theasperformance
determined of
bythe
DST: (a) SOCKalman
cubature estimation; and (CKF)
filtering (b) SOC estimation
and error.
CPF algorithms
in the DST test.
Table 1. Comparison of the performance
EstimationofMethod
the cubature CKF
Kalman filtering
CPF (CKF) and CPF algorithms
in the DST test. Estimation Method CKF CPF
Mean absolute error 0.010 0.005
Mean absolute error 0.010 0.005
Maximum error
Estimation 0.038 CPF
0.026
MaximumMethod
error CKF
0.038 0.026
Mean absolute error 0.010 0.005
Maximum error 0.038 0.026

Figure 7. Voltage estimation under DST: (a) voltage; and (b) voltage error.
Figure 7. Voltage estimation under DST: (a) voltage; and (b) voltage error.

To compareFigure 7. Voltage estimation

the convergence under
rate of the CPFDST:
and(a)CKF
voltage; and (b) voltage
algorithms, error.value of the SOC
the initial
To compare
was thesignificantly
reset to 0.5, convergencedifferent
rate of the
fromCPF
the and CKF algorithms,
experimental the initial
SOC values. SOC was
value of theresults
The experimental
resetare To compare
toshown
0.5, in Figurethe8.convergence
significantly different
The rate of
from
red solid-line thethe
and CPF and
experimental
green CKFSOC
solid-line algorithms,
are values. the
Theinitial
the estimated value
experimental
results of CPF
of the the SOC
results
and are
was
shown reset to
EKFinmethods. 0.5,
Figure 8.The significantly
Theconvergence different
red solid-line rateand from
green
of the the experimental
CKFsolid-line
algorithmare SOC
wasthe values.
estimated
faster The experimental
than theresults of the CPF
convergence results
and
rate of theEKF
are shown
CPF algorithmin Figure 8. The
in the early red solid-line
stage of SOC and green solid-line are the estimated results of the CPF and
methods. The convergence rate of the CKFestimation
algorithmbut was then the than
faster convergence rate of CKF
the convergence algorithm
rate of the CPF
EKF methods.
became The convergence rateCPF
of the CKF algorithm was faster than when
the convergence rate of the
algorithm inslower
the early than thatof of
stage SOCthe algorithm.
estimation but thenIntersection occurs
the convergence rate of the
CKFestimated
algorithmerrorbecame
CPF algorithm
approaches in therefore,
zero, the early stage
we of conclude
can SOC estimation
that butconvergence
these then the convergence
rates are rate of CKF algorithm
approximate.
slower than that of the CPF algorithm. Intersection occurs when the estimated error approaches zero,
became slower than that of the CPF algorithm. Intersection occurs when the estimated error
therefore, we can
approaches conclude
zero, therefore,that
wethese convergence
can conclude ratesconvergence
that these are approximate.
rates are approximate.

Figure 8. Comparison of convergence rates between the CKF and CPF algorithms.

Figure 8. Driving
4.2.2. New European Comparison
Cycleof convergence rates between the CKF and CPF algorithms.
Operating Conditions
Figure 8. Comparison of convergence rates between the CKF and CPF algorithms.
4.2.2.During actual SOC
New European estimation,
Driving there will be
Cycle Operating noise in the data due to measurement errorss of the
Conditions
current sensor and voltage sensor and interference
4.2.2. New European Driving Cycle Operating Conditions from the measurement environment. This
During actual SOC estimation, there will be noise in
increases the requirements for the robustness of the algorithm. the data due
Thus, wetonext
measurement
tested NEDCerrorss of the
conditions
Duringsensor
current SOCvoltage
actual and estimation,
sensorthere
andwill be noise in
interference the the
from datameasurement
due to measurement errorss
environment. of the
This
increases
current sensorthe requirements
and for the
voltage sensor robustness
and of thefrom
interference algorithm. Thus, we nextenvironment.
the measurement tested NEDC conditions
This increases
the requirements for the robustness of the algorithm. Thus, we next tested NEDC conditions to assess
Energies 2017,10,
Energies2017, 10,457
457 11
11 of
of 15
15
Energies 2017, 10, 457 11 of 15

to assess the robustness of the algorithm. The current profile and magnified profile for NEDC are
the robustness
to assess of the algorithm.
the robustness The current
of the algorithm. profile
The andprofile
current magnified profile for profile
and magnified NEDC forare NEDC
shown are
in
shown in Figure 9.
Figure 9.
shown in Figure 9.

(a) (b)
(a) (b)
Figure 9. New European driving cycle (NEDC): (a) current profile and (b) magnified plot of the
Figure profile.
current 9. New European driving cycle (NEDC): (a) current profile and (b) magnified plot of the
Figure 9. New European driving cycle (NEDC): (a) current profile and (b) magnified plot of the
current profile.
current profile.
The 0.03 V voltage random noise and the 0.1 A current random noise can be controlled by the
circuitThe 0.03 Vthe
board voltage random noise and the 0.1 A current random noise can be controlled by the
The
circuit 0.03 in
board Vin the
actual
voltage
actual
processnoise
random
process
are added
are and the
added
to 0.1
to
verify
verify
the robustness
A current
the randomof
robustness of
the algorithm.
noise
the can The redby
be controlled
algorithm. The red
solid-
the
solid-
line and
circuit green
board in solid-line
the actual are the estimated
process added results ofthe
therobustness
CPF and EKF methods in Figures
The red10–12.
line When
and green
0.03 solid-line
V voltage are the are
random estimated
noise
to
was
verify
results
added,of the
the CPF
average
of the
and EKF algorithm.
methods
absolute error inand
Figures
the
solid-line
10–12.
maximum
and green
When solid-line
0.03 are the random
V voltage estimatednoise
results
was ofadded,
the CPFthe and EKF methods
average absolute inerror
Figuresand 10–12.
the maximum
errorWhenof SOC 0.03 estimation
V voltage by the noise
random CKF was method
added, were
the 1.5%
averageandabsolute
6.8%, respectively,
error and the as presented
maximum errorin
error
Figure of10 SOC
and estimation
Table 2. by
The the CKF
estimation method
accuracy were
of 1.5%
the and
CPF 6.8%,
method respectively,
was lower, as
0.5% presented
and 2.9%, in
SOC estimation
ofFigure 10 and by the
Table 2. CKF
The method
estimationwereaccuracy
1.5% andof6.8%, the respectively,
CPF method aswaspresented
lower, in0.5%
Figureand 102.9%,
and
respectively.
Table Figure 11 and
2. The estimation Tableof3 the
showed that when thelower,
absolute
0.5%value of therespectively.
current random noise
respectively.
was less than Figure
0.1A, 11accuracy
the and Table
average absolute
CPF method
3 showed thatand
error whenwas
the the absolute
maximum
and 2.9%,
value
error of ofSOC
the current
estimation
Figure
random
using
11
noise
the
and Table
was method 3 showed
less thanwere that
0.1A,0.7% when
the and
average the absolute
absolute value
error and of the
and 0.6% current
the maximum random
error noise was
of SOC for less
estimationthan 0.1A,
using the
the
CKF
average absolute error and 3.9%,
the respectively,
maximum error of SOC and 2.6%,using
estimation respectively,
the CKF the CPF
method weremethod.
0.7%
CKF
As method
shown in were
Figure 0.7% and
12andand 3.9%,
Table respectively, andabsolute
0.6% anderror 2.6%,and
respectively, for the CPF method.
and
As 3.9%,
shown respectively,
in Figure 12 and0.6% and4,
Table 4,
the average
2.6%,
the respectively,
average for the
absolute CPFand
error
the maximum
method.
the As shownerror
maximum error
of SOC
in Figure
of CPF12
SOC
estimation
and Table using
4, the the
average CKF method
absolute were
error 1.1%
and theand 4.8%,
maximum respectively,
error of SOCand the accuracy
estimation using of the
the CKF
estimation
method wasusing
better the
than CKFthatmethod
of the CKFweremethod,
1.1% and 0.5%4.8%,
andrespectively,
2.9%, and theThe
respectively. accuracy
results of
showthe that
CPF
method
method were 1.1%
wasabsolute and
better than 4.8%,
that respectively,
ofmaximum and
the CKF method, the accuracy of
0.5% estimation the CPF method
and 2.9%, respectively. was better than that of
the
the average
CKF method, 0.5% error
and and
2.9%, errorThe
respectively. of SOCresults show using
that the the CPFThe
average
results
method
absolute
show
were
error
that
lower
and
the average
than absolute
that oferror
the CKF error and
method. maximum
Overall, error of SOC
we conclude estimation
that the robustnessusing the CPF method were lower
maximum
than that of the of SOC
CKF estimation
method. Overall,using
we the CPF method
conclude that the were lower of
robustness than
of
thethat
the
CPFof
CPF
method
the CKF
method
is superior
is method.
superior
to that ofwe
Overall, theconclude
CKF method. that the robustness of the CPF method is superior to that of the CKF method.
to that of the CKF method.

Figure 10.
Figure 10. 0.03 V voltage random noise
noise experiment
experiment results:
results: (a) SOC;
SOC; and
and (b)
(b) SOC
SOC error.
error.
Figure 10. 0.03 V voltage random noise experiment results: (a) SOC; and (b) SOC error.
Energies 2017, 10, 457 12 of 15
Energies 2017, 10, 457 12 of 15
Energies 2017, 10, 457 12 of 15
Table 2. Performance
Table 2. Performance of
of the
the CKF
CKF and
and CPF
CPF algorithms
algorithms under
under NEDC
NEDC conditions
conditions with
with 0.03
0.03 V
V voltage
voltage
random
Table 2.noise.
Performance
random noise. of the CKF and CPF algorithms under NEDC conditions with 0.03 V voltage
random noise.
Estimation Method CKF
Estimation Method CKF CPF
CPF
Estimation
Mean Method
absolute
Mean absolute CKF
error 0.015
error 0.015 CPF
0.005
0.005
Mean absolute
Maximum error
error 0.015
0.068 0.005
0.029
Maximum error 0.068 0.029
Maximum error 0.068 0.029

Figure 11. 0.1 A current random noise experiment results.

Figure 11. 0.1 A current random noise experiment results.
Table 3.
Table 3. Performance
Performance of
of the
the CKF
CKF and
and CPF
CPF algorithms
algorithms under
under the
the NEDC
NEDC conditions
conditions with
with 0.1
0.1 A
A current
current
Table 3. noise.
random Performance of the CKF and CPF algorithms under the NEDC conditions with 0.1 A current
random noise.
random noise.
Estimation Method CKF CPF
Estimation Method CKF CPF
Estimation
Mean Method
absolute CKF 0.006
error 0.007 CPF
Mean
Mean absolute
absolute error 0.007
error 0.039 0.006
0.007 0.026
0.006
Maximum error
Maximum error
Maximum error 0.039 0.026
0.039 0.026

Figure 12. 0.03 V voltage random noise and 0.1 A current random noise experiment results.
Figure 12.
Figure 12. 0.03
0.03 V
V voltage
voltage random
random noise
noise and
and 0.1
0.1 A
A current
current random
random noise
noise experiment
experiment results.
results.
Energies 2017, 10, 457 13 of 15

Table 4. Performance of the CKF and CPF algorithms under NEDC conditions with 0.03 V voltage
random noise and 0.1 A current random noise.

Estimation Method CKF CPF

Mean absolute error 0.011 0.005
Maximum error 0.048 0.029

5. Conclusions
The CPF algorithm was proposed for the SOC estimation of lithium-ion batteries in electric
vehicles. In this method, preprocessing of the particles was performed using the CKF method and
the suggested density function was generated using the PF algorithm. This CPF approach improved
the estimation accuracy and robustness of the CKF algorithm. The second-order equivalent circuit
was selected as a model to approximate the experimental characteristics of the battery, achieving
the trade-off between computational complexity and precision. The experimental results were fitted
by polynomials to test the parameters {Uoc (SOC ) and R0 , R1 , R2 , C1 , C2 } of the second-order RC
equivalent circuit model. We utilized the DST condition to verify the accuracy of the algorithm, added
noise to the NEDC condition to verify the robustness of the algorithm, and compared the CPF method
with the CKF method. We found that the CPF algorithm exhibited better accuracy and robustness than
the original CKF method. Therefore, the CPF algorithm is more suitable for SOC estimation.

Acknowledgments: This work was supported by the Shenzhen Science and Technology Project
(No. JCY20150331151358137).
Author Contributions: Bizhong Xia and Zhen Sun conceived and designed the experiments; Ruifeng Zhang
performed the experiments; Ruifeng Zhang and Zizhou Lao analyzed the data; Bizhong Xia contributed
reagents/materials/analysis tools; and Zhen Sun wrote the paper.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Zheng, Y.; Ouyang, M.; Li, X.; Lu, L.; Li, J.; Zhou, L.; Zhang, Z. Recording frequency optimization for massive
battery data storage in battery management systems. Appl. Energy 2016, 183, 380–389. [CrossRef]
2. Basu, S.; Hariharan, K.S.; Kolake, S.M.; Song, T.; Sohn, D.K.; Yeo, T. Coupled electrochemical thermal
modelling of a novel Li-ion battery pack thermal management system. Appl. Energy 2016, 181, 1–13.
[CrossRef]
3. Ng, K.S.; Moo, C.-S.; Chen, Y.-P.; Hsieh, Y.-C. Enhanced coulomb counting method for estimating
state-of-charge and state-of-health of lithium-ion batteries. Appl. Energy 2009, 86, 1506–1511. [CrossRef]
4. Zhang, H.; Zhao, L.; Chen, Y. A lossy counting-based state of charge estimation method and its application
to electric vehicles. Energies 2015, 8, 13811–13828. [CrossRef]
5. Feng, F.; Lu, R.G.; Zhu, C.B. A combined state of charge estimation method for lithium-ion batteries used in
a wide ambient temperature range. Energies 2014, 7, 3004–3032. [CrossRef]
6. Lee, S.; Kim, J.; Lee, J.; Cho, B.H. State-of-charge and capacity estimation of lithium-ion battery using a new
open-circuit voltage versus state-of-charge. J. Power Sources 2008, 185, 1367–1373. [CrossRef]
7. Dang, X.; Yan, L.; Xu, K.; Wu, X.; Jiang, H.; Sun, H. Open-circuit voltage-based state of charge estimation of
lithium-ion battery using dual neural network fusion battery model. Electrochim. Acta 2016, 188, 356–366.
[CrossRef]
8. Barbarisi, O.; Vasca, F.; Glielmo, L. State of charge Kalman filter estimator for automotive batteries.
Control Eng. Pract. 2006, 14, 267–275. [CrossRef]
9. Sepasi, S.; Roose, L.R.; Matsuura, M.M. Extended Kalman filter with a fuzzy method for accurate battery
pack state of charge estimation. Energies 2015, 8, 5217–5233. [CrossRef]
10. Tulsyan, A.; Tsai, Y.; Gopaluni, R.B.; Braatz, R.D. State-of-charge estimation in lithium-ion batteries: A particle
filter approach. J. Power Sources 2016, 331, 208–223. [CrossRef]
Energies 2017, 10, 457 14 of 15

11. Dai, H.; Zhu, L.; Zhu, J.; Wei, X.; Sun, Z. Adaptive Kalman filtering based internal temperature estimation
with an equivalent electrical network thermal model for hard-cased batteries. J. Power Sources 2015, 293,
351–365. [CrossRef]
12. Tian, Y.; Xia, B.; Sun, W.; Xu, Z.; Zheng, W. A modified model based state of charge estimation of power
lithium-ion batteries using unscented Kalman filter. J. Power Sources 2014, 270, 619–626. [CrossRef]
13. Shi, P.; Zhao, Y.W.; Shi, P. Application of unscented Kalman filter in the SOC estimation of Li-ion battery
for Autonomous Mobile Robot. In Proceedings of the 2006 IEEE International Conference on Information
Acquisition, Weihai, China, 20–23 August 2006; pp. 1279–1283.
14. Fei, Z.; Guangjun, L.; Lijin, F. Battery state estimation using unscented Kalman filter. In Proceedings of the
2009 IEEE International Conference on Robotics and Automation (ICRA), Kobe, Japan, 12–17 May 2009;
pp. 1863–1868.
15. Zhiwei, H.; Mingyu, G.; Caisheng, W.; Leyi, W.; Yuanyuan, L. Adaptive state of charge estimation for Li-ion
batteries based on an unscented kalman filter with an enhanced battery model. Energies 2013, 6, 4134–4151.
16. Xia, B.Z.; Wang, H.Q.; Wang, M.W.; Sun, W.; Xu, Z.H.; Lai, Y.Z. A new method for state of charge estimation of
lithium-ion battery based on strong tracking cubature Kalman filter. Energies 2015, 8, 13458–13472. [CrossRef]
17. Shen, Y.Q. Hybrid unscented particle filter based state-of-charge determination for lead-acid batteries. Energy
2014, 74, 795–803. [CrossRef]
18. Rumelhart, D.E.; Mcclelland, J.L.; Group, T.P. Parallel distributed processing: Explorations in the
microstructure of cognition foundations. Language 1986, 63, 12–21.
19. Kalman, R.E. , A new approach to linear filtering and prediction problems. J. Fluids Eng. 1960, 82, 35–45.
[CrossRef]
20. Xiong, B.; Zhao, J.; Wei, Z.; Skyllas-Kazacos, M. Extended Kalman filter method for state of charge estimation
of vanadium redox flow battery using thermal-dependent electrical model. J. Power Sources 2014, 262, 50–61.
[CrossRef]
21. Wang, S.; Fernandez, C.; Shang, L.; Li, Z.; Li, J. Online state of charge estimation for the aerial lithium-ion
battery packs based on the improved extended Kalman filter method. J. Energy Storage 2017, 9, 69–83.
[CrossRef]
22. Wan, E.A.; van der Merwe, R. The unscented Kalman filter for nonlinear estimation. In Proceedings of the
IEEE 2000 Adaptive Systems for Signal Processing, Communications, and Control Symposium (Cat. No.
00EX373), Lake Louise, AB, Canada, 4 October 2000; pp. 153–158.
23. Mazaheri, A.; Radan, A. Performance evaluation of nonlinear Kalman filtering techniques in low speed
brushless DC motors driven sensor-less positioning systems. Control Eng. Pract. 2017, 60, 148–156. [CrossRef]
24. Ye, M.; Guo, H.; Cao, B. A model-based adaptive state of charge estimator for a lithium-ion battery using an
improved adaptive particle filter. Appl. Energy 2017, 190, 740–748. [CrossRef]
25. Wang, Y.; Zhang, C.; Chen, Z. A method for state-of-charge estimation of LiFePO4 batteries at dynamic
currents and temperatures using particle filter. J. Power Sources 2015, 279, 306–311. [CrossRef]
26. Xia, B.Z.; Wang, H.Q.; Tian, Y.; Wang, M.W.; Sun, W.; Xu, Z.H. State of charge estimation of lithium-ion
batteries using an adaptive cubature Kalman filter. Energies 2015, 8, 5916–5936. [CrossRef]
27. Sun, T.; Xin, M. Hypersonic entry vehicle state estimation using nonlinearity-based adaptive cubature
Kalman filters. Acta Astronaut. 2017, 134, 221–230. [CrossRef]
28. Zhang, Y.; Huang, Y.; Li, N.; Zhao, L. Embedded cubature Kalman filter with adaptive setting of free
parameter. Signal Process. 2015, 114, 112–116. [CrossRef]
29. Zheng, L.; Zhang, L.; Zhu, J.; Wang, G.; Jiang, J. Co-estimation of state-of-charge, capacity and resistance
for lithium-ion batteries based on a high-fidelity electrochemical model. Appl. Energy 2016, 180, 424–434.
[CrossRef]
30. Li, J.; Wang, L.; Lyu, C.; Wang, H.; Liu, X. New method for parameter estimation of an
electrochemical-thermal coupling model for LiCoO2 battery. J. Power Sources 2016, 307, 220–230. [CrossRef]
31. Jiang, J.; Ruan, H.; Sun, B.; Zhang, W.; Gao, W.; Wang, L.Y.; Zhang, L. A reduced low-temperature
electro-thermal coupled model for lithium-ion batteries. Appl. Energy 2016, 177, 804–816. [CrossRef]
32. He, W.; Williard, N.; Chen, C.; Pecht, M. State of charge estimation for electric vehicle batteries using
unscented kalman filtering. Microelectron. Reliab. 2013, 53, 840–847. [CrossRef]
33. Johnson, V.H. Battery performance models in ADVISOR. J. Power Sources 2002, 110, 321–329. [CrossRef]
Energies 2017, 10, 457 15 of 15

34. Chen, X.; Shen, W.; Cao, Z.; Kapoor, A. A novel approach for state of charge estimation based on adaptive
switching gain sliding mode observer in electric vehicles. J. Power Sources 2014, 246, 667–678. [CrossRef]
35. Sepasi, S.; Ghorbani, R.; Liaw, B.Y. A novel on-board state-of-charge estimation method for aged Li-ion
batteries based on model adaptive extended Kalman filter. J. Power Sources 2014, 245, 337–344. [CrossRef]
36. Lee, J.; Nam, O.; Cho, B.H. Li-ion battery SOC estimation method based on the reduced order extended
Kalman filtering. J. Power Sources 2007, 174, 9–15. [CrossRef]
37. Jossen, A.; Späth, V.; Döring, H.; Garche, J. Reliable battery operation—A challenge for the battery
management system. J. Power Sources 1999, 84, 283–286. [CrossRef]
38. Doucet, A.; Godsill, S.; Andrieu, C. On sequential Monte Carlo sampling methods for Bayesian filtering.
Stat. Comput. 2000, 10, 197–208. [CrossRef]
39. Chen, Q.Y.; Jiang, J.C.; Ruan, H.J.; Zhang, C.P. A New double sliding mode observer for EV lithium battery
SOC estimation. Math. Probl. Eng. 2016, 2016, 8048905. [CrossRef]

© 2017 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).