BioGecko Vol 12 Issue 02 2023

ISSN NO: 2230-5807

State of Charge Estimation of Li-ion Battery Using Unscented Kalman Filter

Sonu Kumar1, S. Koteswara Rao2
1
Post Doctoral Research Fellow, Department of Electronics and Communication Engineering,
Koneru Lakshmaiah Education Foundation, Vaddeswaram, Guntur, Andhra Pradesh-522502, India

2
Professor, Department of Electronics and Communication Engineering, Koneru Lakshmaiah Education
Foundation, Vaddeswaram, Guntur, Andhra Pradesh-522502, India

Abstract
An essential indicator for Li-ion batteries is their state of charge (SOC). The safe operation of a Li-ion
battery can be ensured by an accurate SOC estimation. A precise estimation of the remaining energy level
or SOC of the cell or battery pack is necessary for battery management systems (BMS). However, in an
operational environment with Gaussian noise, the conventional estimation approach, has a cumulative
inaccuracy and is unable to sustain satisfactory results for an extended period of time. The estimation error
brought on by Gaussian noise can be eliminated using the Kalman filtering [Link] the purpose of
estimating the state of charge, three Kalman filters, Extended Kalman Kalman filters (EKF), Adaptive
Extended Kalman filters (AEKF) andUnscented Kalman filters (UKF) were created and studied. The test
system is created in MATALAB/Simulink to examine the effectiveness of the various approaches.
Different models are developed and [Link] findings demonstrate that the suggested UKF based
method outperforms conventional methods and has a higher estimation accuracy under various operating
conditions.A comparison between EKF, AEKF and UKF shows UKF gives minimum SOC estimation error
within the range of 0.10%.

Keywords: Battery; SOC Estimation; Kalman filters; UKF; Dynamic modelling; battery management
system

1. Introduction
Two major issues facing the world in recent years are environmental pollution and the energy crisis [1].
The sharp rise in the number of fossil fuel vehicles makes these issues worse. Sustainable development
places a strong emphasis on the replacement of fossil fuels with low-carbon and effective energy sources.
Energy Storage is a crucial component in producing renewable energy, and Li-ion batteries are the most
popular option due to a number of benefits, including comparably better charge and discharge performance,
high energy and charge [Link] to its high energy density, extended cycle life, and rapid charging
time, lithium-ion batteries are frequently used in smart home applications [2]. Li-ion batteries play a vital
role in the advancement of modern technology, such as electric vehicles (EVs), because to their attributes
of long cycle life, high power endurance and high power density [3-5]. The battery's capacity is reflected
by the state of charge (SOC), which is a crucial indicator. The battery management system's ability to
accurately estimate SOC during the charge and discharge is crucial.
Precise SOC calculation takes into account important data including battery performance and remaining
life [6], which in turn helps to manage and use battery power and energy [7] more effectively. Additionally,
over-discharging and over-charging of the battery, which shorten battery life, cause explosions or flames,
accelerate ageing and permanently harm the cell structure of batteries, which can be controlled using proper
SOC estimation [8].The Battery Management System (BMS), which controls the energy flow in a battery
pack with respect to individual cell voltages, temperature, state of charge, and condition of health, is
typically programmed with the SOC estimation algorithm. The primary purpose of BMS is to keep the
battery system's operating environment secure and to guard against damage [9]. Although the calculation

2596
A Journal for New Zealand Herpetology
 BioGecko Vol 12 Issue 02 2023
ISSN NO: 2230-5807

of battery SOC is a crucial BMS function, the non-linear, intricate electrochemical process in the battery
makes it difficult to estimate accurately online [10].
Many studies have used different techniques to estimate SOC in recent years, including the amperehour
integration approach (Ah) [11], the Kalman filter method (KF) [12–14], and deep reinforcement learning
(DRL) [15–18]. The Ah is used the most frequently among them. The outdated techniques for determining
the SOC include counting Ah and measuring impedance. The SoC can be determined via Ah counting as
given below [19]:
𝑡 𝜂𝐼𝑡
𝑆𝑜𝐶(𝑡) = 𝑆𝑜𝐶(𝑡0 ) + ∫𝑡 3600𝐶0
𝑑𝜏 (1)
0
where SoC(t0), C0, h, and It indicate the battery's starting SOC, maximum capacity, coulomb factor, and
terminal current.
Ah approach, however, has a cumulative inaccuracy and is unable to fix the divergence of the starting value
of SOC [20]. The accuracy decreases over time. The Ah counting method is regarded as imprecise since
the current measurement inaccuracy accumulates. Moreover, it is unable to determine the SoC's starting
value. The second method [21–22] uses the open-circuit voltage (OCV) to calculate SoC in accordance
with the battery's OCV–SoC curve.
The most frequent methods for SOC estimation utilized include: coulomb counting, open circuit voltage
(OCV) estimation, electrochemical impedance spectroscopy (EIS), and filtering. Some of the most
commonly used SOC estimation methods are listed in Table 1. However, the OCV is measured after the
battery has been removed from the circuit, hence this method is unable to identify the SoC when it is
operating continuously. Due to the fact that battery impedance varies with SoC, the SoC can also be inferred
from battery impedance. However, because battery impedance is temperature dependent, additional tools
are required to measure it.
Table 1: Different SOC Estimation Methods
Category Methods
Direct Measurement Open circuit voltage, Terminal voltage,
Impedance method
Book keeping method Coulomb counting, Modified coulomb
counting
Indirect measurement Neural network, support vector, fuzzy logic,
Kalman filter, Extended Kalman filter,
Unscented Kalman filter, Cubature Kalman
filter, Particle filter, H infinity filter,
Nonlinear observer, Sliding mode observer,
Proportional integralobserver
Hybrid methods Coulomb counting and Kalman filter,
Kalman filter and Long short-termmemory

Recently, SOC has been estimated using a variety of adaptive techniques, such as fuzzy logic, neural
networks, adaptive observers, and Kalman filters [23]. Other techniques for SoC estimation include robust
and adaptive observers [24,25]. For a linear or piecewise linear battery model, robust H∞ observers can be
created to estimate the SoC [26]. An H∞ observer taking into account an electrochemical impedance model
for SoC estimate is introduced by Chen et al. [27]. However, the robust-observer approach cannot be used
in low-cost microelectromechanical (MEMS) devices because the H∞ algorithm requires extensive matrix
operations.
For SoC estimation, adaptive model reference observer [28], particle filter [29] and nonlinear approaches
[30] are also used. Hu et al. [31] provided a technique for estimating SoC taking into account the time-

2597
A Journal for New Zealand Herpetology
 BioGecko Vol 12 Issue 02 2023
ISSN NO: 2230-5807

varying model parameters taking into account temperature variations during the tests. Nevertheless,
determining model parameters for various temperature ranges takes time and is expensive.
Another nonlinear technique for SoC estimation is sliding-mode observer (SMO). This observer takes the
battery's nonlinear model with uncertainties into account [32, 33]. The chattering phenomenon is the SMO-
based approaches' fundamental flaw. These observers have strong robustness because they can deal with
the ambiguity in the model parameters. In order to minimize chattering, Chen et al. [34] introduce an
adaptive switching-gain SMO for SoC estimation. The strategy is challenging to put into practise, though.
Another group of SoC estimate techniques is intelligence-based techniques, which includes neural
networks, fuzzy neural networks, and fuzzy adaptive neural networks [35–39]. These techniques have
various drawbacks. Although the neural network-based approaches do not require a mathematical model
of the batteries, complete and trustworthy datasets are required for training, testing and validation.
Inaccurate SoC estimation can also result from using the same training dataset for batteries of varied ages
[40]. Additionally, this method's implementation on the CPU presents its own challenges. The fuzzy
inference systems also rely on the expertise of experts [41]. Problems arise while choosing the membership
functions for the fuzzy system design due to expert disagreements. There are also learning-based options
for the SoC estimation, including deep learning [42] and machine learning [43]. The distribution of the
training data and test data is assumed to be the same by the deep-learning algorithms for the SoC estimate.
But in reality, this presumption is incorrect. A deep-transfer neural network with multiscale distribution
adaptation was given in [44] for the SoC estimate as a solution to this issue, but the main disadvantage of
this approach was its difficult implementation. Since they do not require the dynamic model of the battery,
learning-based approaches are often good for SoC estimate. However, a sizable, trustworthy training
dataset is needed for these techniques. Additionally, they need the distribution of the training and test data
as well as pricey graphics processing units (GPU).
Rudolph Kalman [45] created one of the most used algorithms, the KF, in 1960. It was initially used to
predict the trajectories of both manned and unmanned spacecraft. The Kalman filter, a recursive algorithm
for estimating state variables of a dynamic system, has been used to predict battery SoC. Despite the
measurement noise, this optimal observer can offer a precise evaluation of the states. Among these
techniques, the Kalman filtering seems to be very promising [46]. Plett [47–49] presented a technique in
2004 for using the KF to estimate the SOC of LICs, which is not observable directly. This approach was
frequently modified in later works, leading to a variety of KF-based state estimation implementations. In
real-world applications, it filters the system's input and output signals to precisely forecast the dynamic
state of the system. The interference brought on by white noise in the system can be eliminated by the
extended Kalman filter (EKF) approach, and the cumulative error brought on by the ampere-time
integration method can be reduced.
As the battery model is nonlinear, the SoC is commonly estimated using the extended Kalman filter (EKF)
[50–53]. Additionally, the lifetime of lithium-ion batteries has changed their electrochemical
characteristics, which can cause the EKF method to estimate SoC and SoH incorrectly. KFs have the
advantage of taking model and measurement errors into account, which produces a robust estimating
behaviour. A cell depends on SOC, temperature, current and age [54], and because a model cannot account
for every scenario, its uncertainties change as it operates.
In [55], an adaptive extended Kalman filter (AEKF), which employs a covariance adaptation technique, is
presented to enhance the performance of the Kalman filter in SoC estimation. The estimation algorithm's
linearization inaccuracy is this filter's fundamental flaw. In [56,57], the unscented Kalman filter (UKF) and
adaptive UKF are used to address this issue. Additionally [41], implements a sigma-point EKF on the
battery model. Recently, an interactive multi-model UKF [59] and a central-difference Kalman filter
(CDKF) were also constructed for SoC estimation. These solutions, however do not account for model
uncertainties and rely heavily on precise battery models.
However, the main drawbacks of the existing approaches can be summed up as (a) the traditional observers,
like SMO and H∞, have chattering in their response or require complicated mathematical calculation and

2598
A Journal for New Zealand Herpetology
 BioGecko Vol 12 Issue 02 2023
ISSN NO: 2230-5807

not able to estimate SOC when measurements noise are present, (b) the artificial intelligence-based
approaches require a trustworthy training dataset or specialists' knowledge of the battery; (c) KF based
approaches depends on precise battery models and not able to estimate SOC properly when model
uncertainties are present. A detailed description about the advantages and disadvantages different SOC
estimation methods are listed in Table 2.
Table 2. Merits and demerits of various SoC estimation techniques
Technique Advantages Disadvantages
Ampere-hour (Ah) counting Less calculation, low cost, and Error accumulates
approach easy implementation
Impedance measurement Low calculation, low cost, and Sensitive to temperature change
approach easy implementation and time-taking process
AI Battery model not required Large and accurate training data
(Deep learning and is necessary, and both training
Machine learning) algorithm and test data must be
distributed, costly GPU
Kalman filters-based approach Can estimate while Battery model accuracy is
measurement and process required, as well as knowledge
noises are present of measurement and process
noises.
H∞ observer SoC estimation without Heavy processors are required
knowledge of the statistical for calculation.
properties of the battery
Sliding mode-based observers Tolerance to uncertainties in the Chattering occurrences and slow
model convergence

This study offers an UKF based SOC estimation utilising an equivalent battery circuit model. The
equivalent circuit model demonstrates the nonlinear relationship between the OCV and SOC by including
capacitors, resistors and a nonlinear voltage source. The battery models are described in the below section.
2. Battery SOC Estimation Model
For use in several applications, electrochemical batteries come in a variety of models. The electro-circuit
model is suitable for the implementation of estimate algorithms like SoC or SoH estimation. These models
can be categorised into electrochemical, electro-circuit and intelligent models. The Rint model [60],
Thevenin model [61], and Rngv model [62] are three common Li-ion battery related models. Among these,
the Thevenin model can instantly represent the Li-ion battery's operational condition without adding too
much delay to track the actual voltage, ensuring the model's correctness over the course of a lengthy
simulation. Fig. 1 depicts the Thevenin model's structural layout.

Fig. 1. Thevenin equivalent circuit model

The circuit equations for Thevenin's model are given below using Kirchhoff's Voltage Law and Kirchhoff's
Current Law as reference.

2599
A Journal for New Zealand Herpetology
 BioGecko Vol 12 Issue 02 2023
ISSN NO: 2230-5807

𝑈𝑃 𝐼𝐿
𝑈̇𝑃 = − +
𝑅𝑃 𝐶𝑃 𝐶𝑃
𝑈𝐿 = 𝑈𝑃 + 𝑈𝑂𝐶 + 𝐼𝐿 𝑅0 (2)
where IL is the current through the resistance R0, UL is terminal voltage for analogous circuit, UP is the
terminal voltage for the polarization capacitor CP, UOC is open circuit voltage of equivalent circuit.
The concept of Ah is used to describe the value of SOC, and the resulting equation is illustrated below.
𝑡
∫𝑡 𝑘𝑇 𝑖(𝜏)𝑑𝜏
0
𝑆𝑂𝐶(𝑡) = 𝑆𝑂𝐶(𝑡0 ) + (3)
𝑄𝑁
where SOC (t0) is the SOC of Li-ion battery at moment t0, kT is Li-ion battery's temperature correction
factor at temperature T,i (τ) is battery's current at instant τ, and QN is its rated capacity.
Calculating the battery's open circuit voltage will yield the SOC. This method's SOC estimate equation is
provided below.
𝑡
∫𝑡 𝑘𝑇 𝑖(𝜏)𝑑𝜏
0
𝑆𝑂𝐶(𝑡) = 𝑆𝑂𝐶(𝑂𝐶𝑉) + (4)
𝑄𝑁
where OCV is the open circuit voltage value, SOC (OCV) is the matching SOC's starting value. The discrete
state space model of this estimating method is depicted below when combined (3) with the Thevenin model.
∆𝑇
𝑆𝑂𝐶 1 0 𝑆𝑂𝐶𝑘−1 𝑄𝑁
[ 𝑈 𝑘] = [ −
∆𝑇 ] [
𝑈 ]+[ −
∆𝑇 ] 𝐼𝑘−1 + 𝑤𝑘−1 (5)
𝑃,𝑘 0 𝑒 𝑅𝑃𝐶𝑃 𝑃,𝑘−1
(1 − 𝑒 𝑅𝑃𝐶𝑃 ) 𝑅𝑃
𝑈𝐿,𝑘 = 𝑂𝐶𝑉 (𝑆𝑂𝐶𝑘 ) + 𝑈𝑃,𝑘 + 𝑅0 𝐼𝑘 + 𝑣𝑘 (6)

where ∆T is the discrete step size, wk-1denotes process noise at instant k -1, vk denotes observed noise at
instant k.
This nonlinear estimating issue can be resolved using the EKF technique. It uses a recursive technique to
accomplish minimum variance estimation and can provide the estimate's error. This algorithm is an optimal
autoregressive data processing algorithm. The discrete nonlinear state space model is given below.
𝑥 = 𝑓(𝑥𝑘−1 , 𝑢𝑘−1 , 𝑤𝑘−1 )
{ 𝑘
𝑦𝑘 = ℎ(𝑥𝑘 , 𝑢𝑘 , 𝑣𝑘 )
𝑤𝑘 ~ (0, 𝑄)
𝑣𝑘 ~ (0, 𝑅) (7)
Process noise at instant k-1 is denoted by wk-1, measurement noise at instant k is denoted by vk, covariance
of wk and vk are denoted by Q and R, respectively.
The system is linearized at an operating point to convert a nonlinear problem into a linear one. At this
moment, the state equation is converted into the Taylor expansion at xk-1 = xk-1|k-1, wk-1 = 0.
̃ 𝑘 ~ (0, 𝐿𝑘 𝑄𝐿𝑇𝑘 )
𝑤 (8)
𝑇
𝑣̃𝑘 ~ (0, 𝑀𝑘 𝑅𝑀𝑘 ) (9)

The EKF method must estimate the Pk and xk in two separate ways during each sample cycle, including
priori estimation and posteriori estimation. The priori estimate is used to determine the Kalman gain
coefficient Kk between the two phases.
The robust-CDKF approach [63] that has been suggested uses the electrical model that is illustrated in
Figure 2. Electrochemical models, experimental models, electrical models, abstract models based on
artificial intelligence and more models exist for electrochemical batteries. Thevenin models, Impedance
models, Runtime based models and Randle equivalent-circuit models are only a few of the several types of
electrical models. All of the battery's dynamic properties, such as the non-linear OCV, current, temperature,

2600
A Journal for New Zealand Herpetology
 BioGecko Vol 12 Issue 02 2023
ISSN NO: 2230-5807

number of cycles, time-dependent storage capacity and transient responses are included in this
[Link] to the Kirchhoff laws, the terminal voltage can be written as follows.
𝑉𝑡 = 𝑉𝑜𝑐 (𝑆𝑂𝐶) − 𝑉1 − 𝑉2 − 𝐼𝑡 𝑅𝑖𝑛 (10)

Fig. 2. The battery model

The dynamics of the state of charge and polarization voltages are as follows:
1 𝐼
̇ =−
𝑆𝑂𝐶 𝑆𝑂𝐶 − ( 𝑡 ) + ∆𝑓1 (11)
𝑅0 𝐶0 𝐶0
𝑉 𝐼
𝑉̇2 = − 𝑅 2𝐶 + 𝐶𝑡 + ∆𝑓3 (12)
2 2 2

Where voltages V1 and V2 are corresponds to the electrochemical and concentrate respectively.
Uncertainties in the battery model and the internal/external disturbances are included in Δf1 to Δf3. The
updated mean covariance matrix and updated mean estimation of the battery's state variables, as well as the
SoC are determined in the last phase.
𝑋̂𝑖 = 𝑋̂𝑖|𝑖−1 + 𝐾𝑖 (𝑌𝑖 − 𝑌̂𝑖 ) (13)

𝑃̂𝑖 = 𝑃̂𝑖|𝑖−1 − 𝐾𝑖 𝑃̂𝑦 𝐾𝑖𝑇 (14)

Varying-parameter model
It is preferred to include the battery's hysteresis as an additional system state. The discussed issues with the
zero-state model are resolved by adding SOC to the state vector to augment the hysteresis and by utilising
a Kalman filtering approach to estimate both OCV and SOC. The hysteresis-state model developed by Plett
[64] is applied in this work using the following formulation:
𝑑ℎ(𝑆𝑂𝐶,𝑡) ̇ − ℎ(𝑆𝑂𝐶, 𝑡))
̇ )(𝑀 (𝑆𝑂𝐶, 𝑆𝑂𝐶)
= 𝛾𝑠𝑔𝑛(𝑆𝑂𝐶 (15)
𝑑𝑆𝑂𝐶
The maximum polarisation brought on by battery hysteresis as a result of SOC and its rate of change is
known as 𝑀(𝑆𝑂𝐶, 𝑆𝑂𝐶 ̇ ).
The model yield output can be represented considering hysteresis.
𝑉𝑘 = OCV(𝑆𝑂𝐶𝑘 ) – 𝑅𝑖𝑘 + ℎ𝑘 (16)
Extended Kalman filtering for zero-state hysteresis model [64] is summarized below.
Nonlinear state-space model
𝜂𝑖 𝑖𝑘 ∆𝑡
𝑆𝑂𝐶𝑘+1 = 𝑆𝑂𝐶𝑘 − 𝐶𝑛
+ 𝑤𝑘 (17)
𝑉𝑘 = OCV(𝑆𝑂𝐶𝑘 ) – 𝑅𝑖𝑘 − 𝑠𝑘 𝑀 + 𝑣𝑘 (18)
Where 𝑤𝑘 is independent zero-mean, gaussian noises with covariance 𝑃𝑣
𝑣𝑘 is independent zero-mean, gaussian noises with covariance 𝑃𝑤
∂OCV(𝑆𝑂𝐶 )
𝐶𝑘̇ = ∂𝑆𝑂𝐶 𝑘 |𝑆𝑂𝐶𝑘 =𝑆𝑂𝐶
̇ 𝑘− (19)
𝑘
Initializing k=0,
𝑆𝑂𝐶0+ = 𝐸[𝑆𝑂𝐶0 ] (20)

2601
A Journal for New Zealand Herpetology
 BioGecko Vol 12 Issue 02 2023
ISSN NO: 2230-5807

𝑃+𝑆𝑂𝐶
̂
̂ 0+ )(𝑆𝑂𝐶0 − 𝑆𝑂𝐶
̇ ,0 =E [(𝑆𝑂𝐶0 − 𝑆𝑂𝐶 ̂ 0+ )𝑇 ] (21)
Computing k=1, 2, …..
𝜂 𝑖𝑘−1 ∆𝑡
̂ 𝑘− = 𝑆𝑂𝐶
𝑆𝑂𝐶 ̂ 𝑘−1
+
- 𝑖𝐶 (22)
𝑛
− +
𝑃𝑆𝑂𝐶
̂ ,𝑘 = 𝑃𝑆𝑂𝐶
̂ ,𝑘−1 + 𝑃𝑤 (23)
Update on measurements
− ̂𝑇 ̂ ̂ − ̂𝑇 −1
𝐿𝑘 = 𝑃𝑆𝑂𝐶
̂ ,𝑘 𝐶𝑘 [𝐶𝑘 𝑃𝑆𝑂𝐶 ,𝑘 𝐶𝑘 + 𝑃𝑣 ] (24)
̂ 𝑘+ = 𝑆𝑂𝐶
𝑆𝑂𝐶 ̂ 𝑘− + 𝐿𝑘 [𝑉𝑘 − 𝑂𝐶𝑉(𝑆𝑂𝐶𝑘 ) + 𝑅𝑖𝑘 + 𝑠𝑘 𝑀] (25)
+ ̂ ̂ −
𝑃𝑆𝑂𝐶
̂ ,𝑘 = (1 − 𝐿𝑘 𝐶𝑘 )𝑃𝑆𝑂𝐶 ,𝑘 (26)

Fig. 3. MATLAB Simulink model of Kalman filtering based SOC Estimation [65]

Fig. 4. Flowchart for the implementation of KF based method

2602
A Journal for New Zealand Herpetology
 BioGecko Vol 12 Issue 02 2023
ISSN NO: 2230-5807

The MATLAB Simulink model for the extended Kalman filter is represented in Figure 3. The flowchart
for implementation of the proposed Kalman filtering based method, is given in Figure 4.

3. Kalman filters implementation

To forecast the state of a physical or actual process, the KF is based on a set of differential equations (a
model). As a result, by adjusting the state variables, it minimises the error between a linear system's
measured and expected output. The filter is frequently used in the battery field to forecast cell voltage using
an ECM and a coulomb counting. The relationship between the SOC and the OCV is taken into
consideration for this reason. The discrepancy between the estimated and measured voltages is then
compared, and by modifying the SOC and other ECM variables, it is made as small as possible. State
estimation for linear systems can be accomplished using a linear Kalman filter [66, 67].
The linear Kalman filter (LKF) is not frequently utilised in literature because of the non-linear cell
behaviour. The KF can be used with batteries using first-order Taylor approximation of the differential
equations to linearize the system and measurement matrices in the actual state. Extended Kalman Filter
(EKF) is the name of this strategy [68–78]. However, because of the linearization error and the neglect of
the higher-order derivatives of the Taylor approximation, filter estimation can lead to erroneous behaviour
and divergence of the filter [78]. The sigma point Kalman filter (SPKF) was created for this purpose. Here,
a set of sigma points are used to approximate the linearization without the need for derivatives [78-80]. The
unscented Kalman filter (UKF) and the central difference Kalman filter (CDKF) are two popular varieties
of the SPKF.
An UKF based on the unscented transformation is provided in [67,70,81-85]. By skipping the creation of
the system and measurement matrices, this transformation can be used to approximatively determine the
desired values and the covariance of a random variables propagated through a nonlinear function [78]. The
interpolation used by Stirling provides the foundation for the CDKF [78,8687]. The derivation is omitted,
just like in the case of the UKF. The application of scaling and gain factors is related to the distinction
between the two filters. The UKF employs three scaling factors, compared to just one in the CDKF.
Both filters drawback is that each time step's square root calculation of the covariance matrix using the
Cholesky factorization is necessary. The positive definition of the covariance matrix cannot be guaranteed,
and rounding errors can happen. Paper [78,82], introduced the square root forms of the UKF and CDKF to
minimize calculation error. The Cholesky factorization is not calculated in each time step but is just updated
in this case. Adaptive EKF (AEKF) were developed for decreasing the time for tuning of filters[88-94]. In
this, the difference between the actual and anticipated output voltage is used to calculate the process and
measurement noise online.

3.a. Kalman filter

Noise is added to the state-space notation in the discrete ECM to account for model and measurement
uncertainty. The process noise and measurement noise are represented respectively by the two random
variables wk ∈ Rn and vk ∈ Rm. With this notation, the measurement equation and the state-space
representation are expanded to:
𝑥𝑘+1 = 𝐴𝑘 𝑥𝑘 + 𝐵𝑘 𝑢𝑘 + 𝑤𝑘 (27)
𝑦𝑘 = 𝐻𝑘 𝑥𝑘 + 𝐷𝑘 𝑢𝑘 + 𝑣𝑘 (28)

If the mean values are zero and the measurement noise and process noise are uncorrelated, it may be
assumed that:
𝐸[𝑤𝑤 𝑡 ] = Q (29)
𝑡
𝐸[𝑣𝑣 ] = r (30)

As a result, E is the statistical expectation operator, r is the measurement noise covariance and Q is the
process noise matrix covariance [95].

2603
A Journal for New Zealand Herpetology
 BioGecko Vol 12 Issue 02 2023
ISSN NO: 2230-5807

Fig. 5. Calculation sequence of a Kalman filter

Considering certain assumptions, the algorithm simplifies to the following calculation sequence.
Initializing
𝑥̂0+ = 𝐸[𝑥0 ] (31)
𝑃0+ = 𝐸[(𝑥0 − 𝑥̂0+ )(𝑥0 − 𝑥̂0+ )𝑡 ] (32)
Predicting
𝑥̂𝑘− = 𝐴𝑘−1 𝑥̂𝑘−1 +
+ 𝐵𝑘−1 𝑢𝑘−1 (33)
𝑃𝑘 = 𝐴𝑘−1 𝑃𝑘−1 𝐴𝑡𝑘−1 + 𝑄
− +
(34)
Correcting gain
𝑦𝑘 = 𝐻𝑘 𝑥̂𝑘− + 𝐷𝑘 𝑢𝑘 (35)
𝐾𝑘 = 𝑃𝑘− 𝐻𝑘𝑡 (𝐻𝑘 𝑃𝑘− 𝐻𝑘𝑡 + 𝑟)−1 (36)
After correcting
𝑥̂𝑘+ = 𝑥̂𝑘− − 𝐾𝑘 (𝑈𝑘 − 𝑦𝑘 ) (37)
𝑃𝑘+ = (𝐼 − 𝐾𝑘 𝐻𝑘 )𝑃𝑘− (38)

This kind of Kalman filter is referred to as a linear Kalman filter (LKF), in the work that follows. All
varieties of Kalman filters use the given calculation process.

3.b. Extended Kalman filter

The state-space equation and measurement equation are taken to have the following forms in order to
account for nonlinear behaviour:
𝑥𝑘+1 = 𝑓(𝑥𝑘 , 𝑢𝑘 ) + 𝑤𝑘 (39)
𝑦𝑘 = 𝑒(𝑥𝑘 , 𝑢𝑘 ) + 𝑣𝑘 (40)

Given f and e are two differentiable functions. The first order Taylor expansion can be used to approximate
these two functions if their time deviation is minimal. Additionally, either xˆ+ k or xˆ- k, depending on the
most recent state approximation, is used to evaluate the functions. This process makes a distinction between
the LKF and the EKF. These presumptions allow the matrices A and H from Equations 27 and 28 to be re-
written in the form:
𝜕𝑓(𝑥𝑘 ,𝑢𝑘 )
𝐴𝑘 = 𝜕𝑥𝑘
|𝑥𝑘 (41)

2604
A Journal for New Zealand Herpetology
 BioGecko Vol 12 Issue 02 2023
ISSN NO: 2230-5807

𝜕𝑒(𝑥𝑘 ,𝑢𝑘 )
𝐻𝑘 = 𝜕𝑥𝑘
|𝑥𝑘 (42)

The equations 41 and 42 are often referred to as Jacobian matrices. The extended Kalman filter method
functions analogous to the linear Kalman filter when these variances are taken into consideration.

3.c. Adaptive Extended Kalman filter

It is possible to apply the KF to non-linear systems because of the EKF. There are some systems, though,
whose dynamical processes and parameters cannot be precisely identified. The KF then makes inaccurate
estimations as a result.
The measurement noise and the process noise can utilise the system's remaining degrees of freedom to find
a solution to this issue. An accurate estimation of the condition and a quick transient response are produced
by selecting these values properly [88,89,96]. The process noise and the measurement noise can also change
in each time step as a result of fluctuating environmental effects (like temperature), which affect the
approximation [76]. In place of using the measurements noise values, which takes average of overall
potential states of the random variables, the concept is to utilise the average values of deviation of measured
and anticipated measurements values at the most recent time step. Despite the fact that this goes against the
Kalman gain's ability to minimise, the KF replacement takes into account the system's actual behaviour.
Calculating the moving average V φ of the measurement deviation can be used to put this into practise:
1
𝑉𝑘∅ = ∅ ∑𝑘𝑚=𝑘−∅+1(𝑈𝑘 − 𝑦𝑘 ) (𝑈𝑘 − 𝑦𝑘 )𝑡 (43)

Using window sizes of φ ≤ k and φ ∈N. The measurements noise and the process noise matrix are updated
as follows [88,89,96] based on the averaged error.
𝑟𝑘 = 𝑉𝑘∅ +𝐻𝑘 𝑃𝑘− 𝐻𝑘𝑡 (44)
∅ 𝑡
𝑄𝑘 = 𝐾𝑘 𝑉𝑘 𝐾𝑘 (45)

3.d. Unscented Kalman filter/ Sigma point Kalman filter

Neither a linear nor a Gaussian distributed system is possible. Additionally, the impact of the noise may
not always be linear. As a result, the measurement equation and the state-space equation are
𝑥𝑘+1 = 𝑓(𝑥𝑘 , 𝑢𝑘 , 𝑤𝑘 ) (46)
𝑦𝑘 = 𝑒(𝑥𝑘 , 𝑢𝑘 , 𝑣𝑘 ) (47)

The random variable is approximated by other vectors using the differentiable functions f and e, where the
covariance matrix and mean value are created so that they are equivalent to the state parameters. The choice
of these vectors is unrestricted because both the weights and the number of each sigma point are arbitrary.
The probability distribution that was created, though, is simply a rough approximation of the Gaussian
distribution. As a result, only in the first two moments of a Taylor approximation are the mean value and
covariance matrix equivalent. All odd moments are zero, which is equivalent to the Gaussian distribution.
Apart from the Gaussian distribution approximation, the UKF algorithm is identical to all other KFs. Given
the process noise, it is necessary to determine 2n+1 sigma points λn, where n is the state vector's length.
The estimation of the following state can therefore be obtained using the state space in Eq. (46). The
measurement equation and covariance matrices are then computed. The derivation and the entire algorithm
are described in [78,80,81].

4. Simulation Results and Analysis

The estimation for SOC was carried out using extended Kalman filtering, adaptive extended Kalman
filtering and unscented Kalman filtering. For this input measured data contains time, voltage, current and
temperature. The simulation results obtained are given in below Figures 6-13.

2605
A Journal for New Zealand Herpetology
 BioGecko Vol 12 Issue 02 2023
ISSN NO: 2230-5807

At first the estimation using extended Kalman filtering was attempted. The Figure 6, shows the measured
and estimated battery SOC using extended Kalman filter. The nominal capacity for the battery under
investigation is 4.81Ah. It was observed that the battery terminal voltage decreased from 4.2V to 3.73V in
4 hours duration.

Fig. 6. Actual and estimated Terminal voltage based on EKF vs time plot

Fig. 7. Terminal voltage error based on EKF vs time plot

The battery terminal voltage error is represented in Figure 7. It was observed that the terminal voltage
error is within 0.02V range. The RMSE (root mean square error) value for the battery terminal voltage is
1.31%.

Fig. 8. Coulomb counting and estimated SOC (%) based on EKF vs time plot
The Actual SOC (theoretically calculated using coulomb counting) and SOC estimated using EKF method
is shown in Figure 8. It can be seen that the actual and estimated SOC is almost matching. The difference
in actual and estimated SOC is also studied. The error is SOC estimation is shown in Figure 9. The root

2606
A Journal for New Zealand Herpetology
 BioGecko Vol 12 Issue 02 2023
ISSN NO: 2230-5807

mean square error (RMSE) value for SOC estimation is 1.94%. Maximum error is SOC estimation was
observed as 2.80%.

Fig. 9. EKF Based Estimated SOC Error (%) vs time plot

Secondly, adaptive extended Kalman filter (AEKF) based method was used to estimate the SOC of the
battery. In AEKF method an additional Kalman gain is included in comparison to EKF method. AEKF
method initially gives more error in comparison to EKF method. However, after few minutes it gives better
result for the SOC estimation. Figure 10 shows the estimated actual and estimated SOC after 0.5 hours.
The estimated SOC error value was also calculated and shown in Figure 11. The RMSE value for SOC
Error using AEKF for entire interval was 4.98%. However, it can be seen in the Figure 11 that, maximum
SOC error after 0.5 hours interval was within 0.65%.

Fig. 10. Actual and Estimated SOC (%) based on AEKF vs time plot

Fig. 11. Estimated SOC error based on AEKF (%) vs time plot

2607
A Journal for New Zealand Herpetology
 BioGecko Vol 12 Issue 02 2023
ISSN NO: 2230-5807

Thirdly, unscented Kalman filter (UKF) based method was used to estimate the SOC of the battery. This
gives comparatively better results than the previous two methods. This simulation was run for greater time
interval i.e. multiple charge and discharge cycles and it gave satisfactory performance. The actual and
estimated SOC based on UKF method is shown in Figure 12.

Fig. 12. Actual and estimated SOC based on UKF (%) vs time plot

Fig. 13. Actual and estimated based on UKF SOC (%) vs time plot
The difference between the actual SOC and SOC estimated using UKF method is shown in Figure 13. The
above result shows that the error is in the range of less than 1%. The best result was obtained in UKF
method by tuning the covariance matrices. The best result obtained using this method has RMSE value for
SOC Estimation as 0.10%. Maximum SOC error was observed as 0.51%. Hence, it can be said that the
UKF gives better estimate for SOC than EKF and AEKF.

5. Conclusion
In this paper, different Kalman filters have been used for the state of charge estimation of the Li-ion battery.
These algorithms have been implemented is simulation and the estimation results are compared. Kalman
filtering results were compared with experimental data as well. The measurement data and all Kalman
filtering methods obtained result had good agreement (less than 5% error), showing that the suggested
approaches may accurately predict the SOC of the battery in dynamic [Link] simulation results
show that the SOC estimation accuracy of UKF is better than the other KF methods i.e. EKF and [Link]
simulation result of UKF shows anerror of less than 0.51%.Authors suggest that UKF based methods can
be adopted for SOC estimation in battery management system of Li-ion batteries.

References
[1] S. Pang, et al., Interconnection and damping assignment passivity-based control applied to onboard
DC–DC power converter system supplying constant power load, IEEE Trans. Ind. Appl. 55 (6) (2019)
6476–6485, [Link]

2608
A Journal for New Zealand Herpetology
 BioGecko Vol 12 Issue 02 2023
ISSN NO: 2230-5807

[2] Zubi, G.; Dufo-López, R.; Pardo, N.; Pasaoglu, G. Concept development and techno-economic
assessment for a solar home system using lithium-ion battery for developing regions to provide electricity
for lighting and electronic devices. Energy Convers. Manag. 2016, 122, 439–448.
[3] Y. Liu, R. Ma, S. Pang, L. Xu, D. Zhao, J. Wei, Y. Huangfu, F. Gao, A nonlinear observer SOC
estimation method based on electrochemical model for lithium-ion battery, IEEE Trans. Ind. Appl. 57
(2021) 1094–1104, [Link]
[4] Z. Wei, C. Zou, F. Leng, B.H. Soong, K. Tseng, Online model identification and state-of-charge
estimate for lithium-ion battery with a recursive total least squares-based observer, IEEE Trans. Ind.
Electron. 65 (2) (2018) 1336–1346, [Link]
[5] C. Zhang, L.Y. Wang, X. Li, W. Chen, G.G. Yin, J. Jiang, Robust and adaptive estimation of state of
charge for lithium-ion batteries, IEEE Trans. Ind. Electron. 62 (8) (2015) 4948–4957,
[Link]
[6] Ng KS, Moo CS, Chen YP, et al. (2008) State-of-charge estimation for lead-acid batteries based on
dynamic open-circuit voltage. Power and Energy Conference, 2008. PE Con 2008. IEEE 2nd International
972–976.
[7] Xiong R, Cao J, Yu Q, et al. (2018) Critical review on the battery state of charge estimation methods
for electric vehicles. Ieee Access 6: 1832–1843.
[8] Ren H, Zhao Y, Chen S, et al. (2019) Design and implementation of a battery management system with
active charge balance based on the soc and soh online estimation. Energy 166: 908–917.
[9] Satadru D, Beshah A, Pierluigi P (2015) Nonlinear robust observers for state-of-charge estimation of
lithium-ion cells based on a reduced electrochemical model. IEEE Trans Control Syst Technol 23:
1935–1942.
[10] Tang X, Liu B, Gao F (2017) State of charge estimation of lifepo4 battery based on a gainclassifier
observer. Energy Procedia 105: 2071–2076.
[11] M. Kwak, B. Lkhagvasuren, J. Park, J. You, Parameter identification and SOC estimation of a battery
under the hysteresis effect, IEEE Trans. Ind. Electron. 67 (11) (2020) 9758–9767,
[Link]
[12] S. Afshar, K. Morris, A. Khajepour, State-of-charge estimation using an EKF-based adaptive observer,
IEEE Trans. Control Syst. Technol. 27 (5) (2019) 1907–1923,
[Link]
[13] Q. Wang, J. Kang, Z. Tan, M. Luo, An online method to simultaneously identify the parameters and
estimate states for lithium ion batteries, Electrochim. Acta 289 (2018) 376–388,
[Link]
[14] Bo Yang, Junting Wang, Pulin Cao, Tianjiao Zhu, Hongchun Shu, Jiao Chen, Jin Zhang, Jiawei
Zhu, Classification, summarization and perspectives on state-of-charge estimation of lithium-ion batteries
used in electric vehicles: A critical comprehensive survey, J. Energy Storage 39 (2021) 102572,
[Link]
[15] F. Sanchez Gorostiza, F.M. Gonzalez-Longatt, Deep reinforcement learning-based controller for SOC
management of multi-electrical energy storage system, IEEE Trans. Smart Grid 11 (6) (2020) 5039–5050,
[Link]
[16] Guozhou Zhang, Weihao Hu, Di Cao, Qi Huang, Jianbo Yi, Zhe Chen, FredeBlaabjerg, Deep
reinforcement learning based approach for proportional resonance power system stabilizer to prevent ultra-
low-frequency oscillations, IEEE Trans. Smart Grid 11 (6) (2020) 5260–5272,
[Link]
[17] Guozhou Zhang, Weihao Hu, Di Cao, Wen Liu, Rui Huang, Qi Huang, Zhe Chen, FredeBlaabjerg,
Data-driven optimal energy management for a wind-solar-diesel-battery-reverse osmosis hybrid energy
system using a deep reinforcement learning approach, Energy Convers. Manage. 227 (2021) 113608,
[Link]

2609
A Journal for New Zealand Herpetology
 BioGecko Vol 12 Issue 02 2023
ISSN NO: 2230-5807

[18] D. Cao, et al., Reinforcement learning and its applications in modern power and energy systems: A
review, J. Mod. Power Syst. Clean Energy 8 (6) (2020) 1029–1042,
[Link]
[19] Rezaei, O.; Alinejad, M.; Nejati, S.A.; Chong, B. An optimized adaptive estimation of state of charge
for Lithium-ion battery based on sliding mode observer for electric vehicle application. In Proceedings of
the 2020 8th International Conference on Intelligent and Advanced Systems (ICIAS), Kuching, Malaysia,
13–15 July 2021; pp. 1–6.
[20] M. Shehab El Din, A.A. Hussein, M.F. Abdel-Hafez, Improved battery SOC estimation accuracy using
a modified UKF with an adaptive cell model under real EV operating conditions, IEEE Trans
TranspElectrif 4 (2) (2018) 408–417, [Link]
[21] Pei, L.; Lu, R.; Zhu, C. Relaxation model of the open-circuit voltage for state-of-charge estimation in
lithium-ion batteries. IET Electr. Syst. Transp. 2013, 3, 112–117.
[22] Cheng, M.W.; Lee, Y.S.; Liu, M.; Sun, C.C. State-of-charge estimation with aging effect and
correction for lithium-ion battery. IET Electr. Syst. Transp. 2015, 5, 70–76.
[23] N. Watrin, B. Blunier, A. Miraoui, Transportation Electrification Conference and Expo (ITEC), 2012
IEEE, pp.1-6, 18-20 June 2012.
[24] Gholizadeh, M.; Yazdizadeh, A. State of charge estimation of a lithium-ion battery using robust
nonlinear observer approach. IET Electr. Syst. Transp. 2018, 9, 1–7.
[25] Zhang, Z.; Zhou, D.; Xiong, N.; Zhu, Q. Non-fragile H¥ nonlinear observer for state of charge
estimation of lithium-ion battery based on a fractional-order model. Energies 2021, 14, 4771.
[26] Ebrahimi, F.; Abedi, M. Design of a robust central difference Kalman filter in the presence of
uncertainties and unknown measurement errors. Signal Process. 2020, 172, 107533.
[27] Chen, N.; Zhang, P.; Dai, J.; Gui, W. Estimating the state-of-charge of lithium-ion battery using an H-
infinity observer based on electrochemical impedance model. IEEE Access 2020, 8, 26872– 26884.
[28] Verbrugge, M.; Tate, E. Adaptive state of charge algorithm for nickel metal hydride batteries including
hysteresis phenomena. J. Power Sources 2004, 126, 236–249.
[29] Chen, Z.; Sun, H.; Dong, G.; Wei, J.; Wu, J.I. Particle filter-based state-of-charge estimation and
remaining-dischargeable-time prediction method for lithium ion batteries. J. Power Source 2019, 414, 158–
166.
[30] Xia, B.; Chen, C.; Tian, Y.; Sun, W.; Xu, Z.; Zheng, W. A novel method for state of charge estimation
of lithium-ion batteries using a nonlinear observer. J. Power Sources 2014, 270, 359–366.
[31] Hu, Y.; Yurkovich, S. Battery cell state-of-charge estimation using linear parameter varying system
techniques. J. Power Sources 2012, 198, 338–350.
[32] Gholizadeh, M.; Salmasi, F.R. Estimation of state of charge, unknown nonlinearities, and state of
health of a lithium-ion battery based on a comprehensive unobservable model. IEEE Trans. Ind. Electron.
2013, 61, 1335–1344.
[33] Rezaei, O.; Moghaddam, H.A.; Papari, B. A fast sliding-mode-based estimation of state-of charge for
lithium-ion batteries for electric vehicle applications. J. Energy Storage 2022, 45, 103484.
[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.
[35] Chan, C.C.; Lo, E.W.C.; Shen, W. The available capacity computation model based on artificial neural
network for lead–acid batteries in electric vehicles. J. Power Sources 2000, 87, 201–204.
[36] Shen, W.X.; Chan, C.C.; Lo, E.W.C.; Chau, K.T. A new battery available capacity indicator for electric
vehicles using neural network. Energy Convers. Manag. 2002, 43, 817–826.
[37] Yang, F.; Li, W.; Li, C.; Miao, Q. State-of-charge estimation of lithium-ion batteries based on gated
recurrent neural network. Energy 2019, 175, 66–75.
[38] Habibifar, R.; Karimi, M.R.; Ranjbar, H.; Ehsan, M. Economically based distributed battery energy
storage systems planning in microgrids. In Proceedings of the Iranian Conference on Electrical Engineering
(ICEE), Mashhad, Iran, 8–10 May 2018; pp. 1257–1263.

2610
A Journal for New Zealand Herpetology
 BioGecko Vol 12 Issue 02 2023
ISSN NO: 2230-5807

[39] Charkhgard, M.; Farrokhi, M. State-of-charge estimation for lithium-ion batteries using neural
networks and EKF. IEEE Trans. Ind. Electron. 2010, 57, 4178–4187.
[40] Xu, P.; Liu, B.; Hu, X.; Ouyang, T.; Chen, N. State-of-charge estimation for lithium-ion batteries
based on fuzzy information granulation and asymmetric Gaussian membership function. IEEE Trans. Ind.
Electron. 2021, 69, 6635–6644.
[41] Huang, Z.; Fang, Y.; Xu, J. SoC estimation of li-ion battery based on improved EKF algorithm. Int. J.
Automot. Technol. 2021, 22, 335–340.
[42] Hannan, M.A.; How, D.N.; Lipu, M.S.; Mansor, M.; Ker, P.J.; Dong, Z.Y.; Sahari, K.S.M.; Tiong,
S.K.; Muttaqi, K.M.; Mahlia, T.M.; et al. Deep learning approach towards accurate state of charge
estimation for lithium-ion batteries using self-supervised transformer model. Sci. Rep. 2021, 11, 1–13.
[43] Lipu, M.S.H.; Hannan, M.A.; Hussain, A.; Saad, M.H.; Ayob, A.; Uddin, M.N. Extreme learning
machine model for state-of-charge estimation of lithium-ion battery using gravitational search algorithm.
IEEE Trans. Ind. Appl. 2019, 55, 4225–4234.
[44] Bian, C.; Yang, S.; Miao, Q. Cross-domain state-of-charge estimation of Li-ion batteries based on
deep transfer neural network with multiscale distribution adaptation. IEEE Trans. Transp. Electrif.
2020, 7, 1260–1270.
[45] Ng, K.S.; Moo, C.; Chen, Y.; Hsieh, Y. Enhanced coulomb counting method for estimating stateof-
charge and state-of-health of lithium-ion batteries. Appl. Energy 2009, 86, 1506–1511,
doi:10.1016/[Link].2008.11.021.
[46] W. He, N. Williard, C. Chen, M. Pecht, Microelectronic Reliability (2013). Available online at:
[Link]
[47] Jeong, Y.M. Enhanced coulomb counting method with adaptive SOC reset time for estimating OCV.
In IEEE Energy Conversion Congress and Exposition ECCE, Proceedings of the 2014 Energy Conversion
Congress and Exposition, Pittsburgh, PA, USA, 14–18 September 2014; IEEE: New York, NY, USA,
2014; pp. 4313–4318.
[48] Yu, Z.; Huai, R.; Xiao, L. State-of-charge estimation for lithium-ion batteries using a Kalman filter
based on local linearization. Energies 2015, 8, 7854–7873.
[49] He, H. State-of-charge estimation of the lithium-ion battery using an adaptive extended Kalman filter
based on an improved Thevenin model. IEEE Trans. Veh. Technol. 2011, 60, 1461–1469.
[50] Li, W.; Yang, Y.; Wang, D.; Yin, S. The multi-innovation extended Kalman filter algorithm for battery
SOC estimation. Ionics 2020, 26, 6145–6156.
[51] Malysz, P.; Gu, R.; Ye, J.; Yang, H.; Emadi, A. State-of-charge and state-of-health estimation with
state constraints and current sensor bias correction for electrified powertrain vehicle batteries. IET Electr.
Syst. Transp. 2016, 6, 136–144.
[52] Gholizadeh, M.; Yazdizadeh, A. Systematic mixed adaptive observer and EKF approach to estimate
SOC and SOH of lithium–ion battery. IET Electr. Syst. Transp. 2019, 10, 135–143.
[53] Kim, J.; Cho, B.H. State-of-charge estimation and state-of-health prediction of a Li-ion degraded
battery based on an EKF combined with a per-unit system. IEEE Trans. Veh. Technol. 2011, 60, 4249–
4260.
[54] Chen, Z.; Fu, Y.; Mi, C.C. State of charge estimation of lithium-ion batteries in electric drive vehicles
using extended Kalman filtering. IEEE Trans. Veh. Technol 2013, 62, 1020–1030.
[55] Li, J.; Ye, M.; Gao, K.; Jiao, S.; Xu, X. A novel battery state estimation model based on unscented
Kalman filter. Ionics 2021, 27, 2673–2683.
[56] Sun, F.; Hu, X.; Zou, Y.; Li, S. Adaptive unscented Kalman filtering for state of charge estimation of
a lithium-ion battery for electric vehicles. Energy 2011, 36, 3531–3540.
[57] Plett, G. Sigma-point Kalman filtering for battery management systems of LIPB-based HEV battery
packs: Part 2: Simultaneous state and parameter estimation. J. Power Source 2006, 161, 1369– 1384.

2611
A Journal for New Zealand Herpetology
 BioGecko Vol 12 Issue 02 2023
ISSN NO: 2230-5807

[58] Sangwan, V.; Kumar, R.; Rathore, A.K. State-of-charge estimation for Li-ion battery using extended
Kalman filter (EKF) and central difference Kalman filter (CDKF). In Proceedings of the 2017 IEEE
Industry Applications Society Annual Meeting, Cincinnati, OH, USA, 1–5 October 2017; pp. 1–6.
[59] Chen, D.; Wang, C.; Zhu, Z.; Zou, Z. Lithium battery state-of-charge estimation based on interactive
multi-model unscented Kalman filter Algorithm. Energy Storage Sci. Technol. 2020, 9, 257.
[60] V.H. Johnson, Battery performance models in ADVISOR, J Power Sources 110 (2002) 321–329,
[Link]
[61] J. Kim, B.H. Cho, State-of-charge estimation and state-of-health prediction of a li-ion degraded battery
based on an EKF combined with a per-unit system, IEEE Trans. Veh. Technol. 60 (9) (2011) 4249–4260,
[Link]
[62] Z. Chen, Y. Fu, C.C. Mi, State of charge estimation of lithium-ion batteries in electric drive vehicles
using extended Kalman filtering, IEEE Trans. Veh. Technol. 62 (3) (2013) 1020–1030,
[Link]
[63] Chen, M.; Rincon-Mora, G.A. Accurate electrical battery model capable of predicting runtime and IV
performance. IEEE Trans. Energy Convers. 2006, 21, 504–511.
[64] G. Plett, J. Power Sources 134 (2004) 262-276.
[65] I.U Khalil, Matiullah Ahsan, Irshad Ullah, Arbab Adnan, Naveed Khan, Shah Nawaz, SOC prediction
of Lithium-Ion Battery using Extended Kalman Filter, IEEE Xplore.
[66] K. Lim, H. A. Bastawrous, V.-H. Duong, K. W. See, P. Zhang, and S. X. Dou. “Fading Kalman filter-
based real-time state of charge estimation in LiFePO4 battery-powered electric vehicles”. In: Appl. Energy.
169 (2016), pp. 40–48. doi: 10.1016/[Link].2016.01.096.
[67] D. Andre, C. Appel, T. Soczka-Guth, and D. U. Sauer. “Advanced mathematical methods of SOC and
SOH estimation for lithium-ion batteries”. In: J. Power Sources 224 (2013), pp. 20–27. doi:
10.1016/[Link].2012.10.001.
[68] G. L. Plett. “Extended Kalman filtering for battery management systems of LiPBbased HEV battery
packs Part 3. State and parameter estimation”. In: J. Power Sources 134.2 (2004), pp. 277– 292. doi:
10.1016/[Link].2004.02.033.
[69] P. Zarchan and H. Musoff. Fundamentals of Kalman filtering: A practical approach. 3rd ed. Vol. 232.
Progress in astronautics and aeronautics. Reston and Va: American Institute of Aeronautics and
Astronautics, 2009.
[70] Y. He, X. Liu, C. Zhang, and Z. Chen. “A new model for State-of-Charge (SOC) estimation for high-
power Li-ion batteries”. In: Appl. Energy. 101 (2013), pp. 808–814. doi:
10.1016/[Link].2012.08.031.
[71] J. Lee, O. Nam, and B. H. Cho. “Li-ion battery SOC estimation method based on the reduced order
extended Kalman filtering”. In: J. Power Sources 174.1 (2007), pp. 9–15. doi:
10.1016/[Link].2007.03.072.
[72] S. Lee, J. Kim, J. Lee, and B. H. Cho. “State-of-charge and capacity estimation of lithium-ion battery
using a new open-circuit voltage versus state-of-charge”. In: J. Power Sources 185.2 (2008), pp. 1367–
1373. doi: 10.1016/[Link].2008.08.103.
[73] J. Kim and B. H. Cho. “State-of-Charge Estimation and State-of-Health Prediction of a Li-Ion
Degraded Battery Based on an EKF Combined With a Per-Unit System”. In: IEEE T VehTechnol 60.9
(2011), pp. 4249–4260. doi: 10.1109/TVT.2011.2168987.
[74] H. He, H. Qin, X. Sun, and Y. Shui. “Comparison Study on the Battery SoC Estimation with EKF and
UKF Algorithms”. In: Energies 6.10 (2013), pp. 5088–5100. doi: 10.3390/en6105088.
[75] J. T. B. A. Kessels, B. Rosca, H. J. Bergveld, and van den Bosch,P. P. J. “On-line battery identification
for electric driving range prediction”. In: Vehicle Power and Propulsion Conference (2011), pp. 1–6. doi:
10.1109/VPPC.2011.6043022.

2612
A Journal for New Zealand Herpetology
 BioGecko Vol 12 Issue 02 2023
ISSN NO: 2230-5807

[76] C. Zhang, J. Jiang, W. Zhang, and S. M. Sharkh. “Estimation of State of Charge of Lithium-Ion
Batteries Used in HEV Using Robust Extended Kalman Filtering”. In: Energies 5.12 (2012), pp. 1098–
1115. doi: 10.3390/en5041098.
[77] J. Li, J. Klee Barillas, C. Guenther, and M. A. Danzer. “A comparative study of state of charge
estimation algorithms for LiFePO4 batteries used in electric vehicles”. In: J. Power Sources 230 (2013),
pp. 244–250. doi: 10.1016/[Link].2012.12.057.
[78] Rudolph van der Merwe. “Sigma-Point Kalman Filters for Probabilistic Inference in Dynamic State-
Space Models”. PhD thesis. Oregon: Oregon Health & Science University, 2004.
[79] G. L. Plett. “Sigma-point Kalman filtering for battery management systems of LiPBbased HEV battery
packs Part 2: Simultaneous state and parameter estimation”. In: J. Power Sources 161.2 (2006), pp. 1369–
1384. doi: 10.1016/[Link].2006.06.004.
[80] G. L. Plett. “Sigma-point Kalman filtering for battery management systems of LiPB based HEV
battery packs Part 1: Introduction and state estimation”. In: J. Power Sources 161.2 (2006), pp. 1356– 1368.
doi: 10.1016/[Link].2006.06.003.
[81] H. He, R. Xiong, and J. Peng. “Real-time estimation of battery state-of-charge with unscented
Kalman filter and RTOS uCOS-II platform”. In: Appl. Energy. (2015). doi:
10.1016/[Link].2015.01.120.
[82] Y. Tian, B. Xia, W. Sun, Z. Xu, and W. Zheng. “A modified model based state of charge estimation
of power lithium-ion batteries using unscented Kalman filter”. In: J. Power Sources 270 (2014), pp. 619–
626. doi: 10.1016/[Link].2014.07.143.
[83] E. A. Wan and van der Merwe, Rudolph. “The unscented Kalman Filter”. In: Kalman filtering and
neural networks. Ed. by S. S. Haykin. Adaptive and learning systems for signal processing,
communications, and control. New York: Wiley, 2001, pp. 221–280.
[84] Y. Xing, W. He, M. Pecht, and K. L. Tsui. “State of charge estimation of lithium-ion batteries using
the open-circuit voltage at various ambient temperatures”. In: Appl. Energy. 113 (2014), pp. 106–115. doi:
10.1016/[Link].2013.07.008.
[85] X. Zheng and H. Fang. “An integrated unscented kalman filter and relevance vector regression
approach for lithium-ion battery remaining useful life and short-term capacity prediction”. In: Rliabengsyst
safe 144 (2015), pp. 74–82. doi: 10.1016/[Link].2015.07.013.
[86] L. Wang, L. Wang, C. Liao, and J. Liu. “Sigma-point Kalman filter application on estimating battery
SOC”. In: IEEE Vehicle Power and Propulsion Conference (VPPC). 2009, pp. 1592–1595. doi:
10.1109/VPPC.2009.5289604.
[87] J. Remmlinger, M. Buchholz, T. Soczka-Guth, and K. Dietmayer. “On-board state-ofhealth monitoring
of lithium-ion batteries using linear parameter-varying models”. In: J. Power Sources 239 (2013), pp. 689–
695. doi: 10.1016/[Link].2012.11.102.
[88] J. Han, D. Kim, and M. Sunwoo. “State-of-charge estimation of lead-acid batteries using an adaptive
extended Kalman filter”. In: J. Power Sources 188.2 (2009), pp. 606–612. doi:
10.1016/[Link].2008.11.143.
[89] H. He, R. Xiong, and H. Guo. “Online estimation of model parameters and state-ofcharge of LiFePO4
batteries in electric vehicles”. In: Appl. Energy. 89.1 (2012), pp. 413–420. doi:
10.1016/[Link].2011.08.005.
[90] R. Xiong, H. He, F. Sun, X. Liu, and Z. Liu. “Model-based state of charge and peak power capability
joint estimation of lithium-ion battery in plug-in hybrid electric vehicles”. In: J. Power Sources 229 (2013),
pp. 159–169. doi: 10.1016/[Link].2012.12.003.
[91] F. Sun, R. Xiong, and H. He. “Estimation of state-of-charge and state-of-power capability of lithium-
ion battery considering varying health conditions”. In: J. Power Sources 259 (2014), pp. 166– 176. doi:
10.1016/[Link].2014.02.095.
[92] H. He, Z. Liu, and Y. Hua. “Adaptive Extended Kalman Filter Based Fault Detection and Isolation for
a Lithium-Ion Battery Pack”. In: Energy Procedia 75 (2015), pp. 1950–1955. doi:

2613
A Journal for New Zealand Herpetology
 BioGecko Vol 12 Issue 02 2023
ISSN NO: 2230-5807

10.1016/[Link].2015.07.230.
[93] Y. Wang, C. Zhang, and Z. Chen. “An adaptive remaining energy prediction approach for lithium-ion
batteries in electric vehicles”. In: J. Power Sources 305 (2016), pp. 80–88. doi:
10.1016/[Link].2015.11.087.
[94] R. Xiong, X. Gong, C. C. Mi, and F. Sun. “A robust state-of-charge estimator for multiple types of
lithium-ion batteries using adaptive extended Kalman filter”. In: J. Power Sources 243 (2013), pp. 805–
816. doi: 10.1016/[Link].2013.06.076.
[95] G. L. Plett. “Extended Kalman filtering for battery management systems of LiPB based HEV battery
packs Part 1. Background”. In: J. Power Sources 134.2 (2004), pp. 252–261. doi:
10.1016/[Link].2004.02.031.
[96] R. Xiong, F. Sun, X. Gong, and H. He. “Adaptive state of charge estimator for lithiumion cells series
battery pack in electric vehicles”. In: J. Power Sources 242 (2013), pp. 699–713. doi:
10.1016/[Link].2013.05.071.

2614
A Journal for New Zealand Herpetology