ELEC-8900 Special Topics:

Advanced Energy Storage Systems

Lecture 06: Battery Modeling
Instructor: Dr. Balakumar Balasingam
June 16, 2022

Contents
1 Introduction 2

2 Elements of Electrical Equivalent Circuit Models 3

2.1 DC Equivalent Circuit Model . . . . . . . . . . . . . . . . . . . . 4
2.2 AC Equivalent Circuit Model . . . . . . . . . . . . . . . . . . . . 5

3 Reduced Order Models 7

3.1 Ideal Battery Model . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.2 Open Circuit Voltage Model . . . . . . . . . . . . . . . . . . . . . 9
3.3 Relaxation Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.4 Hysteresis Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.5 Enhanced Self Correcting Model . . . . . . . . . . . . . . . . . . 17
3.6 R-int Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.7 Other Reduced Order Models . . . . . . . . . . . . . . . . . . . . 19

4 Battery Power 19

5 Battery Capacity 21
5.1 Total Capacity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
5.2 Discharge Capacity . . . . . . . . . . . . . . . . . . . . . . . . . . 23
5.3 Rated Capacity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
5.4 Custom Defined Capacity . . . . . . . . . . . . . . . . . . . . . . 24

6 State of Health 26

7 Battery Packs 27

8 Battery Simulator 29

9 Summary 30

1
1 Introduction
Battery modelling is a necessary step in BMS development in which we try
to represent the functioning of a battery through mathematical models. Two
important functions of a battery are charging and discharging. In addition to
that, a battery goes through other functionalities, such as aging. A good battery
model will represent all these different functionalities that happen within a
battery cell.
Before we get into battery modelling, let us take a brief look at the inter-
nal structure and workings of a Li-ion battery. Figure 1 shows the four main
components of a Li-ion rechargeable battery: anode, cathode, separator and
electrolyte.

Anode (-) Cathode (+)

Carbon Separator LiCoO2

Negative Positive
current current
collector collector
(Copper) (Aluminum)

Li-ion
Electrolyte

Figure 1: Rechargeable Lithium ion battery cell.

The negative electrode (anode) delivers electrons to the circuit during dis-
charge and it receives electrons during charge. In most Li-ion cells the anode
is composed of graphite; a hexagonal structure made of 6 carbon atoms that
are tightly bonded together to form one layer called graphene. Graphene layers
are loosely held to each other through weak Van der Waals forces. The Li-ions
intercalates between the graphene layers.
Current collectors take no part in the chemical reaction. They are responsible
for delivering or receiving electrons from the circuit. Their sole purpose is to
help collect more electrons after the reaction happens. As a result, it can help
reduce the internal resistance of the battery. In most Li-ion cells negative current
collectors are made of copper. The positive current collector is usually made of
aluminum.
The positive electrode (cathode) is responsible for receiving electrons from
the circuit during discharge and delivering them during charge. There are multi-
ple chemistries that can be used in the cathode as active material in Li-ion cells,
such as, lithium cobalt oxide, lithium nickel manganese cobalt oxide, lithium

2
manganese oxide, lithium nickel cobalt aluminum oxide, and lithium iron phos-
phate. The principal of operation is the same for all these chemistries where
the material is made of layer-like structure that stays the same through the
discharge and charge process while Li-ions can intercalate between this layered
structure.
A separator is used to prevent any direct contact between the anode and
the cathode that would lead to a short circuit inside the cell. The separator
should also be an electric insulator so that electrons are prevented from traveling
internally between the anode and cathode without going through the external
circuit. It should also allow for the flow of Li-ions, this is done by using a porous
material that has holes that are small enough for the Li-ions to pass but not
the active material.
Finally, the electrolyte is responsible for the flow of Li-ions through the cell
between the anode and cathode. It is comprised of a solvent and salt; the
most common salt used is lithium hexafluorophosphate which is dissolved in the
solvent such as ethylene carbonate or diethyl carbonate.
The underlining principle for Li-ion battery operation during discharging is
that the lithium atoms at the anode turn into a Li-ion by losing an electron,
which is delivered to the circuit, then the Li-ion is free to travel all the way to
the cathode where a Li-ion accepts and electron from the circuit and turn into a
lithium atom again. This process is reversed during charging and Li-ions travel
from the cathode to the anode.
The charging and discharging mechanism of a rechargeable battery is af-
fected by external factors, such as, temperature, usage patterns and age. For
example, at low temperatures the mobility of electrons and Lithium ions are
restricted resulting in high electrical resistance; this resistance is often referred
to as the charge transfer resistance. When the battery is repeatedly charged
and discharged (i.e., cycled) it forms a solid electrolyte interface (SEI); this re-
sistance is known as SEI resistance. The depletion of Lithium ions over time
results in the reduction of battery capacity. In order to capture the change
of battery response over time, it is important to accurately model the battery.
The remainder of this chapter provides the details of electrical equivalent circuit
models.

2 Elements of Electrical Equivalent Circuit Mod-

els
The requirement in battery modelling, with respect to battery management,
is to have a model that allows to observe important states and parameters
pertaining to the state of charge and state of health of the battery based on non-
invasive measurements. Possible non-invasive measurements are voltage across
the battery terminals, current through the battery, and temperature on the
battery surface. Hence, electrical equivalent circuit models are widely adopted
to monitor the battery. It is also possible to non-invasively measure humidity

3
and physical expansion of battery cells. And it is possible to model these extra
measurements to observe various useful states related to the safety and health
of a batter pack. However, the discussion in the present chapter is limited to
electrical equivalent circuit models (ECMs).
In the remainder of this section, two different equivalent circuit models are
presented and discussed: the DC equivalent circuit model and the AC equivalent
circuit model.

2.1 DC Equivalent Circuit Model

Figure 2 shows the DC equivalent circuit model (DC-ECM) of a battery. The
electromotive force of the battery is denoted by EMF, h(k) denotes hysteresis,
R0 , R1 , R2 denote ohmic resistance, and C1 , C2 denote capacitance. The voltage
across the battery terminals and the current through the battery are denoted
by v(k) and i(k), respectively. The current through resistances R1 and R2
are denoted by i1 (k) and i2 (k), respectively. This makes the current through
capacitances C1 and C2 to be i(k) − i1 (k) and i(k) − i2 (k), respectively. Here,
the time is denoted in discrete domain, i.e., i(k) denotes the current measured
at time k. The discussions in this chapter also assumes uniform sampling, i.e.,
the time between the sample k and sample k − 1 is assumed to be the same for
all values of k.

h(k)
R0 i1 (k) R1 i2 (k) R2 i(k)
−
+

+
+ − + − +
EMF C1 C2 v(k)
−
−

Figure 2: DC equivalent circuit model.

Based on the notations described so far, and shown in Figure 2, the measured
voltage across the battery terminals can be written as

v(k) = EMF + h(k) + i(k)R0 + i1 (k)R1 + i2 (k)R2 (1)

It can be formally derived (proof omitted) that the currents i1 (k) and i2 (k) can
be derived in terms of i(k) as follows

i1 (k + 1) = α1 i1 (k) + (1 − α1 )i(k) (2)

i2 (k + 1) = α2 i2 (k) + (1 − α2 )i(k) (3)
∆ ∆
where α1 = e− R1 C1 , α2 = e− R2 C2 and ∆ denotes the sampling time.

4
 The quantities R0 , R1 , R2 , C1 , and C2 in Figure 2 are shown without a time
index (k) indicating that these are the ECM parameters. Hysteresis is shown
with time index (k), indicating that hysteresis is not a constant parameter,
rather, it is a quantity that changes with time. Later, the hysteresis will h(k)
will be introduced as a function of the present and past values of current, the
state of charge of the battery, and hysteresis. It must be noted that the ECM
parameters may change with temperature and state of charge of the battery. The
battery impedance significantly increases at low temperatures and low state of
charge regions. The discussion about the ECM in this chapter assumes that the
temperature is constant.

2.2 AC Equivalent Circuit Model

Figure 3 shows the AC equivalent circuit model of the battery which is also
known in the literature as the adaptive Randles equivalent circuit model (AR-
ECM). In addition to the EMF component, the AR-ECM consists of the fol-
lowing elements: stray inductance L, ohmic resistance RΩ , solid electrolytic
interface (SEI) resistance RSEI , solid electrolytic interface (SEI) capacitance
CSEI , double layer (DL) resistance RDL , double layer (DL) capacitance CDL ,
and Warburg impedance Zw . In this chapter, the units of resistance, induc-
tance, and capacitances are referred to in Ohm (Ω), Henry (H), and Farad (F)
unless otherwise explicitly mentioned.
The AC and DC equivalent models are very similar, since they both represent
the same battery. One significant exception in the AC equivalent circuit model
is the Warburg impedance, denoted as Zw in Figure 3. The Warburg impedance
is written as
r
2
Zw , Zw (jω) = σ (4)
jω
where ω indicates angular frequency. It can be noticed that the Warburg
impedance is significant only at very low frequencies. At high frequencies, i.e.,
when ω is significantly high, the effect of Warburg impedance becomes insignif-
icant and the AC equivalent circuit starts to resemble very similar to its DC
counterpart.
It is also possible to add the hysteresis component to the AC equivalent
model. Similarly, the inductive component can be added to the DC equivalent
circuit models as well. However, the effect of inductance is only observed in
very high frequencies limiting it to laboratory studies that most often employ
frequency domain analysis in batteries.
Let us denote the voltage across the battery and the current through it as

V (ω) = Vdc + Ṽ (ω) (5)

˜
I(ω) = Idc + I(ω) (6)

where V (ω) and I(ω) denote the voltage and current measurements consisting
˜
of the DC component (Vdc , Idc ) and AC component (Ṽ (ω), I(ω)). The AC

5
 CSEI CDL

L RΩ ˜
I(ω)
+
RSEI RDL Zw
+
EMF Ṽ (ω)
−
−

Figure 3: AC equivalent circuit model.

impedance of the battery can be written as

Ṽ (ω)
Z(jω) = (7)
˜
I(ω)
1 1
= jωL + RΩ + 1 + 1
RSEI + jωCSEI RCT +Zw (jω) + jωCDL
RSEI RCT + Zw (jω)
= jωL + RΩ + +
1 + jωRSEI CSEI 1 + jω (RCT + Zw (jω)) CDL

By measuring the AC impedance across a battery, the AR-ECM model pa-

rameters can be estimated. Such frequency domain impedance analysis is known
as the electrochemical impedance spectroscopy (EIS) [?]. In EIS the Nyquist
plot is made by plotting the real-part of Z(jω) against the imaginary part of
it. All the parameters of the AC equivalent circuit model shown in Figure 3
can be recovered from a Nyquist plot. The following Matlab codes generate the
Nyquist plot for the ECM parameters corresponding to Figure 3.

Example 1.
The parameters of the Adaptive Randles ECM shown in Figure 3 are:
L = 4 × 10−6 H, RΩ = 0.5 Ω, RSEI = 0.1 Ω, CSEI = 0.2 F, RCT = 0.5 Ω,
CDL = 100 F, and σ = 0.005. Generate the Nyquist plot corresponding
to this battery.

The following MATLAB codes will generate the Nyquist plot shown in Fig-
ure 4 based on the AC ECM model parameters provided in Example 1.
% MATLAB demo of Example 1
% file name: `NyquistPlot.m'
% initialize AC-ECM model parameters

R_Omega = 0.5; R_SEI = 0.1; C_SEI = 0.2;

6
R_CT = .5; C_DL = 100; L= 4e-6; Sig = 0.005;
% create frequency array
fmin = .00001; fmax = 1000; frange = [];
for i=1:ceil(log10(fmax/fmin))
fr = fmin:fmin/4:10*fmin;
frange = [frange fr];
fmin = 10*fmin;
end
w = 2*pi*frange; % angular frequency
Zw = Sig*sqrt(2./(j*w)); % Warburg Imp.
Z = j*w*L + R_Omega + ...
1./(1./(R_CT+Zw)+j*w*C_DL)+ ...
1./(1/R_SEI+j*w*C_SEI);
plot(real(Z), -imag(Z)) % Nyquist plot

0.7

0.6

0.5

0.4

0.3

0.2

0.1

-0.1
0.4 0.6 0.8 1 1.2 1.4 1.6 1.8

Figure 4: Nyquist plot.

Remark 1.
It must be noted that both the DC and AC equivalent circuit models in-
troduced so far did not represent one important parameter of a battery: the
battery capacity. In the next section, the role of battery capacity within the
battery ECM will be made clear. Additional discussions on battery capacity
are provided in Section 5.

3 Reduced Order Models

As it is demonstrated through a simulation example in Section 2.2, the AC
equivalent model requires the application of wide ranging frequency to obtain

7
the Nyquist spectrum. Consequently, the applications of AC equivalent circuit
models are limited to laboratory analysis of batteries. Battery management
systems need to operate in real-time with the help of opportunistic measure-
ments of voltage and current. Practical battery management systems often use
reduced order models made of selected elements of the DC equivalent circuit
model to represent the behaviour of the battery. This section provides further
details about each component of the DC equivalent circuit model.

3.1 Ideal Battery Model

An ideal battery would have the following features:

1. Constant voltage source: It indefinitely acts as a constant voltage source

regardless of the amount of Coulombs taken from it.
2. Zero internal resistance: The voltage across the battery terminals remain
the same regardless of the magnitude of the current.

3. Zero hysteresis: There is no hysteresis (memory) effect in the voltage and

current measurements.
Figure 5 shows the equivalent circuit model of an ideal battery. Ideal battery
model is used in circuit analysis courses, however, it has little use in the study
about battery management systems.

i
+
+
EMF v
−
−

Figure 5: Ideal battery model. An ideal battery is one that functions indef-
initely as a constant voltage source, i.e., the measured terminal voltage would
always be v = EMF regardless of the current i.

Contrast to the ideal battery model shown in Figure 5, real world battery
cells exhibit the following features:

1. Limited voltage source. In practice, the battery voltage does not stay at a
constant v = EMF until it empties. As charge, measured in Coulombs, is
taken away from the battery the EMF voltage gradually drops. The EMF
voltage is widely known as the open circuit voltage (OCV) and denoted
by Vo in this book. The OCV is found to have a monotonous relationship
with the state of charge of the battery. As the battery is discharged its
state of charge decreases and so does the OCV.

8
 2. Electrical resistance effect. In addition to state of charge dependent OCV,
the measured terminal voltage is observed to be affected by voltage drop
due to electrical resistance.
3. Hysteresis effect. It was found that the measured voltage across the bat-
tery terminal is a function of previous current values the battery experi-
enced; this phenomenon is also known as the memory effect.
4. Relaxation effect. It was found that the terminal voltage of a battery
gradually recovers to a higher value after discharging and to a lower value
after charging; this phenomenon is known as the relaxation effect.
In the remainder of this section, we will discuss how electrical equivalent circuit
models (ECM) are used to represent the above four behaviours in real world
batteries.

3.2 Open Circuit Voltage Model

The battery is a limited voltage source. The voltage across the terminals, de-
noted EMF in Figure 5, drops as the battery is discharged and increases when
the battery is charged. The state of charge of the battery is defined as
remaining Coulombs (Ah)
SOC = (8)
battery capacity (Ah)
where SOC ∈ [0, 1] and the unit of the remaining Coulombs and the battery
capacity are both in Ampere hours (Ah). When the battery is charged its SOC
increases and when it is discharged its SOC decreases. The SOC of an empty
battery is 0% and that of a fully charged battery is 100%. The SOC is some
times displayed as a percentage in a typical electronic device, such as the one
shown in Figure 6.
The EMF is referred to as the open circuity voltage (OCV) of the battery.
Theoretically, the OCV has a monotonous relationship with the SOC, i.e., as the
SOC increases so does the OCV and vice versa. Figure 7 shows the OCV-SOC
relationship of a typical Li-ion battery cell. When the battery is empty the SOC
is 0 or 0 % and when it is full the SOC is 1 or 100 %. The OCVs corresponding to
an empty and full battery are denoted by OCVmin and OCVmax , respectively.
The OCV-SOC relationship is useful to estimate the state of the charge of a
battery. In simplified form, the SOC estimation approach works as follows:
measure the OCV across the battery terminals, then use the OCV-SOC curve
in Figure 7 to look-up the SOC of the battery. Given that the OCV-SOC
curve is available as a parameterized function, the SOC estimation reduces to a
root-finding problem.
One of the most important functions of a battery management system is
SOC estimation. Hence, a BMS needs to have the parameters of the OCV-SOC
curve. It is possible to derive the parameters of the OCV-SOC curve based
on the exact chemical compositions of the battery cell. However, in practice,
battery management systems use empirical approaches to obtain the OCV-SOC

9
Figure 6: State of charge of a battery as displayed on a smart phone.

parameters. A latter chapter provides details of various empirical approaches

to estimate the parameters of an OCV-SOC curve.
Empirical OCV-SOC modelling approaches try to fit the measured OCV-
SOC data using polynomial functions. Three such models are given below:
• Polynomial model

Vo (s) =p0 + p1 s + · · · + pn sn (9)

• Combined model

Vo (s) = κ0 + κ1 s−1 + κ2 s + κ3 ln(s) + κ4 ln(1 − s) (10)

• Combined+3 model

Vo (s) = k0 + k1 s−1 + k2 s−2 + k3 s−3 + k4 s−4

+k5 s + k6 ln(s) + k7 ln(1 − s) (11)

where s denotes SOC and Vo (s) denotes the corresponding OCV, i.e.,

EMF ≡ OCV , V◦ (s) (12)

and the coeffieiencts {p0 , p1 , . . . , pn }, {κ0 , κ1 , . . . , κ4 }, and {k0 , k1 , . . . , k7 } are

referred to as the OCV-SOC parameters of the polynomial, combined, and com-
bined+3 models, respectively. In this book, the terms ‘OCV-SOC parameters’
and ‘OCV parameters’ are used interchangeably.
The selection of OCV-SOC model, out of the three given in (9)-(11) or from
many other possible models for curve fitting, is left to the BMS designers. Most

10
 4.2

3.8

3.6

3.4

3.2

2.8
0 20 40 60 80 100
State of charge (%)

Figure 7: OCV-SOC model of a typical Li-ion battery.

times, computational and hardware implementation constraints are prioritized

over accuracy of the model. Out of the three models presented (9)-(11), the
combined+3 model is the most advanced in terms of accuracy. The combined+3
model is designed in a way to accommodate long linear portions and sudden
declines usually found in OCV-SOC data. In this chapter, the combined+3
model is used in demonstrations.
Many existing OCV-SOC models, including (10)-(11), use terms such as
1/s, log(s), and log(1 − s), that are not defined at either s = 0 or at s = 1. In
order to avoid numerical instability, a linear scaling approach can be employed.
Figure 8 shows a linear scaling approach that can be used to avoid substituting
s = 0 and s = 1 in (11). The scaling approach maps the SOC domain s ∈ [0, 1]
to s0 ∈ [0 + , 1 − ] in a linear fashion as described in Figure 8. That way s0
is prevented from reaching 0 or 1 and provides stability in computations [1].
Before computing the OCV, the SOC is scaled as

s0 = (1 − 2)s +  (13)

where the value of  needs to be selected based on the model; it was reported
in [1] that  = 0.175 gives optimal results in Combined model and its variants.

Example 2.
The OCV parameters of a Samsung Galaxy S4 battery, according to
the combined+3 model, is k0 = −9.082, k1 = 103.087, k2 = −18.185,
k3 = 2.062, k4 = −0.102, k5 = −76.604, k6 = 141.199, and k7 = −1.117.
These parameters were obtained after linearly scaling with  = 0.175.
Use these parameters to create and OCV-SOC plot similar to the one
shown in Figure 7.

11
 s0 =  Scaled SOC (s0 ) s0 = 1 − 

s=0 True SOC (s) s=1

 

Figure 8: Linear scaling.

An OCV-SOV plot corresponding to the given parameters can be produced

through the following steps:

• Create SOC values spanning the entire region [0, 1] :

s = [0, 0.001, 0.002, . . . , 0.999, 1] (14)

• Scale SOC values in (14) using (13)

• For each of the scaled SOC, compute the corresponding OCV using (11)
and the given parameters
• Plot s (not s0 ) against the compute OCV
The resulting OCV-SOC plot will be very similar to the one shown in Fig-
ure 7. The following Matlab code will produce as OCV-SOC plot using the
parameters given in Example 2.
% MATLAB demo of Example 2
% filename:SampleOCVSOCplot.m
% Creates OCV-SOC plot
% Uses Combined+3 model and scaling

k0 = -9.082; k1 = 103.087; k2 = -18.185;

k3 = 2.062; k4 = -0.102; k5 = -76.604;
k6 = 141.199; k7 = -1.117;
SOC = 0:.0001:1;
epsilon = 0.175;
zs = SOC*(1-2*epsilon) + epsilon;
OCV = k0*ones(size(SOC)) ...
+ k1*(1./zs) + k2*(1./(zs.^2)) ...
+ k3*(1./(zs.^3)) + k4*(1./(zs.^4))...

12
 + k5*(zs) + k6*(log(zs))...
+ k7*(log(1-zs));
plot(SOC, OCV, '-')

3.3 Relaxation Model

The relaxation effect observed in a battery is represented by the resistor and
capacitor elements shown in Figure 2. Measured relaxation effect in a battery
is shown in Figure 9 using real-world experimental data that was obtained as
follows:

3.85
-0.2

3.8 -0.4

-0.6
3.75

-0.8

3.7
-1

3.65 -1.2
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3

(a) (b)

Figure 9: Relaxation effect. Measured voltage across the battery in response

to changing current over a three minute period.

1. First 15 seconds: zero current.

Since there is no current activity, the voltage is expected to remain con-
stant during the first 15 seconds. Here, the battery was sufficiently rested
in order to remove any prior relaxation effects. As expected, the measured
voltage remained at 3.87 V.

2. Next 30 seconds: constant discharge of id = −1.171A.

When the current is applied, the voltage immediately dropped from 3.87 V
to 3.68 V; this drop is due to the resistive elements in the ECM. Next the
current remains constant and the OCV of the battery reduces as a result;
this is observed through the continuous drop in the measured terminal
voltage.

3. Final 135 seconds: zero current.

When the discharge current is stopped, the voltage immediately returns
to higher value. After that, rather than remaining constant, the voltage
keeps increasing. This observation demonstrates the presence of relaxation
effect within the battery.

13
Example 3.
Consider the battery with the OCV parameters given in Example 2.
The DC-ECM parameters of the battery are as follows: R0 = 0.2 Ω,
R1 = 0.1 Ω, C1 = 2 F, R2 = 0.3 Ω, C2 = 5 F. Consider the following 5
second load current profile that is applied to the battery: -40 mA from 0
to 1 seconds, -120 mA from 1 to 2 seconds, -120 mA from 2 to 3 seconds,
-40 mA from 3 to 4 seconds, and -120 mA from 4 to 5 seconds. Assuming
that the above current was sampled every ∆ = 0.01 seconds, compute
the resulting voltage v across the battery terminals under the following
model assumptions.

1. Ideal battery
2. R-int model (for this case, assume R1 , R2 , C1 , C2 are all zero)
3. RC model (for this case, assume R2 , C2 are zero)

4. 2RC model
Assume the same battery OCV parameters given in Example 2, that the
battery is full at the start and that the battery capacity is Q = 1.5 Ah.

1. In ideal battery, the terminal voltage is equal to V◦ (s(k)) where s(k) is

the SOC at time k which can be computed using the following Coulomb
counting equation
∆i(k)
s(k + 1) = s(k) + (15)
3600Q
where i(k) is the current through the battery in Amperes. Once s(k) is
computed, the terminal voltage v(k) = V◦ (s(k)) can be computed using
(11).
2. Here, the terminal voltage is simply
v(k) = V◦ (s(k)) + i(k)R0 (16)

3. In this case, the effect of the first RC element in Figure 2 will be included
as follows
v(k) = V◦ (s(k)) + i(k)R0 + i1 (k)R1 (17)
where i1 (k) is the current through the resistor R1 can be computed recur-
sively using (2).
4. In this case, the effects of both of the first RC elements in Figure 2 will
be included as follows
v(k) = V◦ (s(k)) + i(k)R0 + i1 (k)R1 + i2 (k)R2 (18)

14
 where i2 (k) is the current through the resistor R2 can be computed recur-
sively using (3).

4.2

4.19

Voltage (V) 4.18

4.17

4.16

4.15

4.14
Ideal
R-int
4.13
1RC
2RC
4.12
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Time (s)

Figure 10: Reduced order ECMs. Simulated terminal voltage for four dif-
ferent reduced order ECMs in Example 3.

The following MATLAB codes generate the answers to Example 3. It makes

use of the battery simulator presented in Section 8.

% initialize battery parameters

k0 = -9.082; k1 = 103.087; k2 = -18.185;
k3 = 2.062; k4 = -0.102; k5 = -76.604;
k6 = 141.199; k7 = -1.117;
Kbatt = [k0; k1; k2; k3; k4; k5; k6; k7];
Batt.Kbatt = Kbatt; Batt.Cbatt = 1.5;
Batt.R0 = .2; Batt.R1 = .1;
Batt.C1 = 2; Batt.R2 = .3;
Batt.C2 = 5; Batt.ModelID = 4;
Batt.SOC = 1; Batt.SOCsf = 0.175;
% store current values in an array
delta = 0.01;
T = 0.01:delta:5;
I = -40*ones(1,length(T));
idx = find((T>1&T<2)|(T>3&T<4));
I(idx) = -120; I = I'/1000;
% (1) Ideal battery model
Batt.ModelID = 2;
Batt.R0 = 0;
[V1] = battSIM(I, Batt, delta);
h=figure; hold on

15
plot(T,V1,'LineWidth',2)
% (2) R-int mode1
Batt.R0 = .2;
[V2] = battSIM(I, Batt, delta);
plot(T,V2,'LineWidth',2)
% (3) 1RC model
Batt.ModelID = 3;
[V3] = battSIM(I, Batt, delta);
plot(T,V3,'LineWidth',2)
% (4) 2RC model
Batt.ModelID = 4;
[V4] = battSIM(I, Batt, delta);
plot(T,V4,'LineWidth',2)

3.4 Hysteresis Model

The presence of hysteresis in a battery can be explained through the following
observations
• After applying a discharging current id , the SOC of the battery is com-
puted to be s1 . Hence, the measured terminal voltage v is expected to
return to Vo (s1 ) after the battery fully relaxes. However, in reality, the
terminal voltage v returns to a value less than the expected Vo (s1 ).
• After applying a charging current ic , the SOC of the battery is computed
to be s2 . Hence, the measured terminal voltage v is expected to return
to Vo (s2 ) after the battery fully relaxes. However, in reality, the terminal
voltage v returns to a value higher than the expected Vo (s2 ).

Experiments further concluded that

• The magnitude of hysteresis voltage depends on the magnitude of current
it experienced over time (i.e., it is a function of both current and time)
• The magnitude of hysteresis voltage depended on the preceding SOC val-
ues of the batteries over time (i.e., it is a function of both SOC and time)
Based on experimental studies, the following model was presented in [?] to
explain the behaviour of the hysteresis.
 
ηγik ∆k
h(k + 1) = exp − h(k)
Q
    (19)
ηγik ∆k
+ exp − − 1 sgn(ik )
Q

where k indicates time, η is the charging efficiency of the battery, Q is the

battery capacity in Ah, γ, ∆k is the sampling time, and ik is the current (+ve
for charging and -ve for discharging) through the battery.

16
 Figure 11 shows a simulated hysteresis behaviour in a battery based on the
hysteresis model (19). In Figure 11(a), a battery was excited by two different
current profiles, Profile A and Profile B, and the resulting hysteresis voltage
is shown. Figure 11(b) shows the expected hysteresis when the profiles were
switched in time.
It shows the non-linear relationship and the memory effect of hysteresis.
In Figure 11(b), the same current profile, with different time order, is used to
simulate hysteresis. The resulting hysteresis voltage is found to be different
from the one shown in Figure 11(a). This simulation explains the non-linear
and convoluted nature of the hysteresis effect in batteries.

50 Profile A Profile B 50 Profile B Profile A

0 0

-50 -50
0 10 20 30 40 50 60 0 10 20 30 40 50 60

1 1

0.5 0.5

0 0

-0.5 -0.5

-1 -1
0 10 20 30 40 50 60 0 10 20 30 40 50 60

(a) (b)

Figure 11: Hysteresis voltage for two different current profiles. The
magnitude of the hysteresis voltage is simulated and do not represent actual
values in typical batteries.

To summarize, the more components the model has the more accurate it
becomes. However, increased model complexity implies increased difficulty, in
terms of required data and computational complexity, in identifying the model
parameters. Also, certain type of data is needed to make some model parameters
observable; this will be further illustrated later. In practical battery manage-
ment systems, various ECM approximations are used. These approximations
are generally referred as reduced order models. Some important reduced order
models are discussed in the next three subsections.

3.5 Enhanced Self Correcting Model

The enhanced self correcting model was first introduced in [2] to represent the
unknowns of a battery ECM in the form of a state-space model. The enhanced

17
self correcting model consists of the following process equation
    
s(k + 1) 1 0 0 0 s(k)
i1 (k + 1) 0 α1 0 0   i1 (k)
 
i2 (k + 1) = 0 0 α2
  
0  i2 (k)
h(k + 1) 0 0 0 Ah (k) h(k)
 
Ac 0  
1 − α1 0  i(k)
+1 − α2
 (20)
0  sgn(i(k))
0 Ah (k) − 1
and the following measurement equation
v(k) = Vo (s(k)) + h(k) + i1 (k)R1 + i(k)R0 (21)
 
where Ac = i k ∆k
Q and Ah (k) = exp − ηγiQk ∆k .
Given voltage measurements v(k), the enhance self-correcting model can be
used to estimate the unknown quantities s(k), i1 (k), i2 (k) and h(k) assuming
that the model parameters α1 , α2 , Ac and Ah (k) are known. When the model
parameters are not known, advanced estimation techniques, such as the expec-
tation maximization algorithm, can be used to estimate them.

3.6 R-int Model

The R-int reduced order battery ECM consists of the OCV-SOC model discussed
in Section 3.2 and an internal resistance Rint . Using the R-int model, shown in

Rint i(k)
+
+
EMF = Vo (s(k)) v(k)
−
−

Figure 12: R-int model.

Figure 12, the the voltage across the battery terminals can now be written as
v(k) = Vo (s(k)) + i(k)Rint (22)
where i(k) is the current through the battery. It must be noted that the DC
equivalent circuit model reduces to the R-int model when the the current i(k)
remains constant for long enough time while the hysteresis effect is ignored. One
the capacitors are saturated, one can write
Rint = R0 + R1 + R2 (23)

18
The R-int battery model is widely studied and adopted in BMS literature for its
simplicity. In this book, various aspects of battery management will be discussed
based on R-int with the understanding that the discussion can be generalized
for ECMs that include both hysteresis and relaxation effects.

3.7 Other Reduced Order Models

It is possible to derive many other reduced order models form the DC-ECM
shown in Figure 2. Indeed, the R-int model described in Section 3.6 is derived
by ignoring the hysteresis and relaxation (RC) effects. An RC model is also
widely used in the literature where one RC element instead of two is employed.
In latter chapters several such reduced order models will be used to explain
various concepts. We will use these models with slight abuse of notation. For
example, when an R-int model is used, the resistance will be denoted R0 instead
of Rint . This convention is followed in this book to reduce the number new
notations for ECM parameters.

4 Battery Power
The DC equivalent circuit model can be used to develop an expression of the
available power in a battery. In this section, R-int reduced order model is
considered for the derivations. Deriving an expression for the general DC-ECM
is left as an exercise. The power of a battery at time k can be expressed using
the following equation

P (k) = v(k)i(k) (24)

where P (k) represents the power at time k. When i(k) is positive, the computed
power denotes the input power to the battery and when i(k) is negative the
computed power denotes the output power.
Based on the R-int model, one can write

v(k) = Vo (s(k)) + i(k)R0

v(k) − Vo (s(k)) (25)
i(k) =
R0
Now, the instantaneous power is written as
 
v(k) − Vo (s(k))
P (k) = v(k) (26)
R0

Using the expression for instantaneous power in (26), the maximum allowable
power of a battery during charging can be written as follows
 
Vmax − Vo (s(k))
PIn,max (k) = Vmax (27)
R0

19
where Vmax = OCVmax is the maximum voltage allowed across the battery
terminals. Similarly, the maximum available power during discharging can be
written as
 
Vmin − Vo (s(k))
POut,max (k) = Vmin (28)
R0
where Vmin = OCVmin is the minimum voltage allowed across the battery termi-
nals. One can notice that Vo (s(k)) > Vmin . Hence, maximum discharge power
at a given time k can be written as −POut,max (k).

Example 4.
Consider the battery in Example 2.

1. What will be the open circuit voltage when the SOC of the battery
is 60%?
2. Assuming the battery internal impedance of R0 = 0.2 Ω calculate
the available peak-power in the battery?

3. The battery was connected to a load which drew 300 mA constant

current for 1 hour. Assume that the battery capacity is 1.2 Ah.
Compute the available peak-power in the battery now.
4. Compute the open circuit voltage of the battery when it is full.

The answers to Example 4 are as follows.

1. To find the OCV at 60% SOC (or s = 0.6), we need to submit s = 0.6
in the combined+3 equation (11) for the given OCV parameters. Before
that, scaling must be done as follows:
s0 = (1 − 2)s +  = (1 − 2)0.6 +  = 0.565
Substituting s0 = 0.565 in (11), we get
V◦ (s) = OCV = 3.8723 V

2. According to Figure 7, Vmin occurs when s = 0. The scaled SOC is

s0 = (1 − 2)0 +  = 0.175
by substituting s0 = 0.175 in (11) we get
Vmin = V◦ (s = 0) = 2.886
By substituting in (28)
 
2.886 − 3.8723
POut,max = −2.886 ≈ 14.23 W
0.2
is the (maximum) available power

20
 3. The new SOC of the battery can be computed as follows:
- The initial SOC was s = 0.6
- The amount of Coulombs removed by the load is 0.3 × 1 = 0.3 Ah.
- Removed Coulombs as a ratio of battery capacity is 0.3/1.2 = 0.25
- New SOC is 0.6 − 0.25 = 0.35
The scaled version of the new SOC is

s0 = (1 − 2)s +  = (1 − 2)0.35 +  = 0.4025 (29)

Now, the OCV can be computed to be

V◦ (s) = V◦ (0.4025) = 4.1917 V (30)

By substituting in (28)
 
3.78 − 3.22
POut,max = 3.22 ≈ 12.68 W
0.2

is the new available power

4. When the battery is full, the SOC becomes s = 1. The scaled version of
the new SOC is

s0 = (1 − 2)1 +  = 0.825 (31)

Vmax = V◦ (0.825) = 4.1917

5 Battery Capacity
Battery capacity is measured in Ampere hours (Ah), i.e., a battery of x Ah in
capacity can provide a load current of x A for one hour. This is only possible
in theory. In practice, a battery of strictly x Ah in Capacity cannot provide
x A of load current for one hour due to the presence of internal resistance. In
this section, we will discuss various ways to define battery capacity for practical
applications.
Let us refer to the R-int equivalent circuit model shown in Figure 12 to
understand more about the practical aspects of battery capacity. As indicated
by the OCV-SOC curve in in Figure 7, the EMF voltage of the battery reaches
to OCVmax when the battery is fully charged. As charge is taken away from
the battery, the EMF voltage drops reaching OCVmin when the battery is fully
empty. The change of EMF is not linear against the state of charge of the
battery. It is also important to note that the operational region of the battery
is between OCVmin and OCVmax , i.e., the battery cannot be charged above
OCVmax and neither can it be discharged below OCVmin . Charging a battery

21
above OCVmax may cause thermal runaway – an irreversible process that trig-
gers meltdown and fire. Discharging a battery below OCVmin may permanently
damage the battery. These two constraints are the focus of several battery man-
agement functionalities discussed throughout this book. The cut-off voltages
OCVmax and OCVmin are important factors to understand battery capacity.
The upper voltage threshold of the battery OCVmax serves as the charge
cut-off voltage, i.e., a. battery charger monitors the voltage v at the battery
terminal (see Figure 12) and terminates charging when v → OCVmax . It must
be noted that, due to the internal resistance R0 , the terminal voltage v will
reach OCVmax before the EMF voltage reaches it. In other words, due to
the internal resistance R0 , the battery charger must be switched off before the
battery is completely full (which happens only when EMF reaches OCVmax ).
In order to fully charge a battery, typical Li-ion battery chargers switch to
constant-voltage mode in which the terminal voltage v is kept at a constant
OCVmax . The charging current will gradually decrease to zero and the battery
become fully charged. Figure 13 shows how the voltage, current and SOC change
during constant-current constant-voltage (CCCV) charging. Most practical Li-
ion battery chargers employ CCCV topology.
The lower voltage threshold of the battery OCVmin serves as the discharge
cut-off voltage, i.e., a battery management system (or protection circuit of a
battery) monitors the voltage v at the battery terminal (see Figure 12) and
disconnects the load when v → OCVmin . Similar to before, the internal re-
sistance R0 causes the terminal voltage v to reach OCVmin before the EMF
voltage reaches it, i.e., the battery needs to be shut down before the battery
is completely empty. In other words, the available Coulombs of a battery de-
pends on the voltage drop caused by the current and the internal resistance R0 .
Due to this, different definitions of capacity are introduced in the next three
subsections.

1.5 1 4.25

Current
SOC 0.9
4.2

0.8
4.15
1 0.7

0.6 4.1

0.5 4.05

0.5 0.4
4
0.3

3.95
0.2

0 0.1 3.9
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

(a) (b)

Figure 13: Constant-current constant-voltage (CC-CV) charging. Left:

The current and SOC of the battery while charging. Right: The voltage of the
battery while charging.

22
 Total Capacity
4.2

4.1
Voltage drop
4

3.9

3.8

3.7

3.6

3.5
Discharge Threshold
3.4
Discharge Capacity
3.3
0 20 40 60 80 100
State of charge (%)

Figure 14: Total capacity vs. discharge capacity in a battery.

5.1 Total Capacity

Total battery capacity is the maximum amount of Coulombs that can be dis-
charged from a battery, starting from a fully charged battery corresponding to
its OCV = OCVmax until the battery is fully discharged, i.e., until the battery
terminal reaches OCV = OCVmin using an infinitesimal load current. The total
capacity specifies a theoretical value for the battery capacity. It order to achieve
this, the charger switch to constant voltage mode and wait until (theoretically)
the charging current becomes zero. In practice, battery capacity can be defined
in many other ways to make useful interpretations. A few such definitions are
demonstrated in the next three subsections.

5.2 Discharge Capacity

Discharge capacity indicates the amount of Coulombs that can be discharged
from a battery at a certain discharge rate (load current). For example, consider
a battery of 1.5 Ah in total capacity. It’s discharge capacity could be 1.2 Ah
at 1 A discharge rate or it could be 1 Ah at 2 A discharge rate. The drop in
discharge capacity with higher load current is due to the fact that the voltage
drop, which amounts to i(k)R0 according to the R-int model shown in Figure 12,
increases with the load current. As the voltage drop increases, the terminal
voltage reaches the discharge cut-off before the battery is completely empty.
Figure 14 describes this scenario where the discharge capacity is less than the
total capacity due to the voltage drop during discharging. As the voltage drop
increases, i.e., as the load current increases, the discharge threshold OCVmin
will be met earlier and the battery and the discharge capacity will be lower.

23
5.3 Rated Capacity
Rated capacity is the manufacturer specified discharge capacity of the battery.
As the battery ages, the rated capacity becomes less and less accurate repre-
sentation of the true discharge capacity of the battery. The rated capacity is
accompanied by a discharge rate, e.g., 2 Ah at C-rate, where C-rate is a popular
way of specifying the load current in terms of battery capacity C.

Example 5.
A battery of 10 Ah capacity is discharged at C/10 rate. How long will
it take to completely discharge the battery?

At C/10 rate, the discharge current is 1 A; it will take 10hr to completely

discharge the battery. In general, when a battery is discharged at C/N rate, it
will last for N hours.

5.4 Custom Defined Capacity

It was described earlier that it takes infinite time to fully charge or discharge the
battery. Coincidently, it was discovered by researchers that completely charing
the battery, or completely discharging it, has negative effects on battery health.
Hence, it is beneficial to maintain the battery in such a way that its SOC is
always maintained within a certain interval. For example, a battery whose SOC
is always maintained between 30% and 70% will last longer, and take less time
to charge, compared to a battery whose SOC is maintained between 10% and
90%. Assume that the capacity of the above battery is 10 Ah; then the effective
capacity of the battery is 4 Ah in the former case and 8 Ah in the latter case.
That is, there is a trade-off involved in selecting an effective capacity; selecting
tighter SOC range results in reduced rated capacity but longer battery life and
vice versa. It is also possible to select these SOC values to allow a certain load.
This is elaborated using Example 6.

Example 6.

A battery has 2 Ah in capacity and an internal resistance of R0 = 0.05 Ω.

It is required to charge this battery fast, using a constant current of 2 A
and the highest load is also expected to be 2 A.
(a) Design an effective SOC range that allows the above requirements.
(b) What is the effective capacity of the battery?
(c) How long it will take to fully charge the battery at the given rate?

Figure 15 shows the required voltage drop (of 0.05 ∗ 2 = 0.1 V) during charg-
ing and discharging of the battery. The battery needs to be shut-off when the
voltage drop hits the thresholds: OCVmax during charging and OCVmin during
discharging and the corresponding SOC values form the required SOC range.

24
 Figure 15: Custom battery design.

(a) According to a visual inspection in Figure 15, the SOC range is approxi-
mately [5%, 85%].
(b) Hence, the effective capacity of the battery is 80% of the true capacity, i.e.,
1.6 Ah.
(c) The battery, according to the above design, can be charged from “empty”
to “full” at a constant rate of 2 A. Hence, the charging time is
1.6 Ah
Tc = = 0.8 hr = 48 min.
2A
In example 6, the effective capacity is computed approximately by graph-
ically drawing the voltage drop along with the OCV-SOC curve. Given the
OCV-SOC parameters, the effective capacity can be precisely computed for
given requirements.

Example 7.

Consider the battery in Example 6. Assume that in a different applica-

tion, the current requirement is increased to 4 A. Recompute the effective
SOC range, effective capacity and the charging time.

Figure 16 shows the required voltage drop (of 0.05 ∗ 4 = 0.2 V) during charg-
ing and discharging of the battery. The corresponding SOC values from the
required SOC range.
(a) According to a visual inspection in Figure 16, the SOC range is approxi-
mately [6%, 71%].
(b) Hence, the effective capacity of the battery is 65% of the true capacity, i.e.,
1.3 Ah.
(c) The battery can be charged using a constant current of 2 A. Hence, the

25
 Figure 16: Custom battery design.

charging time is
1.3 Ah
Tc = = 0.325 hr = 19.5 min.
4A

6 State of Health
The state of health (SOH) of the battery is defined in terms of power fade
(PF) and capacity fade (CF). In this section, formal definitions of PF and CF
is defined first. Then, approaches to define SOH based on PF and CF are
discussed.
Due to SEI growth and other internal chemical reactions the battery impedance
increases over time. When the impedance of the battery increases, the output
power decreases; this phenomenon is known as the power fade. Power fade is
formally defined as
 
P (k)
PF = 1 − 100 % (32)
P (0)

where the available power at time k is defined in (26). Here, the time index k
indicates the elapse of life-cycle event, such as time and charge-discharge cycle;
it is assumed that k = 0 denotes the very initial cycle, e.g., brand new battery.
For accurate results, the power P (k) must be computed at the same temperature
and SOC.
As the battery ages, an increasing portion of its composition loses its ability
to store energy resulting in reduced battery capacity over time. It must be noted
that the battery capacity fade starts from the time the battery is manufactures.
Various factors such as deep discharge, full charge, and extreme temperature
conditions aggravate the capacity fade process. The capacity fade is formally

26
defined as
 
Q(k)
CF = 1− 100 % (33)
Q(0)

where Q(0) is the initial capacity of the battery and Q(k) is the capacity at time
k; similar to before, time k could be either cycle number or a calendar time unit.
Even though battery capacity doesn’t fluctuate as wide as the Ohmic resistance,
small changes in capacity may occur against temperature. Hence, capacity fade
must be computed at the same temperature.
In practice, the exact values of the resistance R0 (k) and capacity Q(k) of
the battery at time k are not known — they need to be estimated. Hence, the
PF can CF equations need to be rewritten as
!
P̂ (k)
PF = 1 − 100 % (34)
P (0)
!
Q̂(k)
CF = 1 − 100 % (35)
Q(0)

where P̂ (k) and Q̂(k) are the estimated values of the power and capacity, re-
spectively, at time k computed for a temperature.
The PF and CF measures can be unified into a single measure of SOH in
various ways. Four such SOH definitions are:
1. SOH = PF
2. SOH = CF
3. SOH = 100 − max {PF, CF}
4. SOH = 100 − min {PF, CF}
The first definition above gives more priority to available power whereas the
second approach prioritizes battery capacity. The third and fourth definitions
of the SOH are the most strictest and most relaxed, respectively.
All SOH measures introduced so far need to be computed based on the as-
sumption that initial values, i.e., P (0) and Q(0), are known. The implication
of this assumption is a significant limiting factor in present day battery man-
agement systems. It remains challenging to evaluate the SOH of an arbitrary
battery; such ability will pave the way for safe, efficient, and reliable approaches
to reuse batteries in various second life applications.

7 Battery Packs
Battery packs are formed by connecting several battery cells in parallel and/or
in series according to the power and capacity requirements. When identical
battery cells are connected in series, the voltage across the terminal increases

27
 E R0 E R0 E R0 E R0
−+ −+ −+ −+

(a) Series connection.

Veq
−+ Req

(b) Equivalent circuit.

Figure 17: Equivalent circuit model of series connection.

R0 R0 R0 R0 Req

+ + + + +
E− E− E− E− E−

(a) Parallel connection. (b) Equivalent

circuit.

Figure 18: Equivalent circuit model of parallel connection.

proportional to the number of cells whereas as the battery capacity remains the
same. Figure 17 shows series connection of four identical battery cells. Here,
the voltage, internal resistance, and the capacity of the resulting pack are

Veq = 4E, R0,eq = 4R0 , Qeq = Q (36)

where Q is the capacity of the individual cell.

When identical battery cells are connected in parallel, the capacity of the
resulting battery pack increases proportional to the number of cells whereas the
voltage remains the same. Figure 18 shows parallel connection of four identical
battery cells. Here, the voltage, internal resistance, and the capacity of the
resulting pack are

Veq = E, R0,eq = R0 /4, Qeq = 4Q (37)

When battery cells are connected in parallel, the voltage across each cell
in the pack is maintained at the same level by the natural flow of electrons.
When the cells are connected in series, the voltage across each cell in the pack
may not remain the same due to various factors. This phenomenon is known as

28
cell imbalance which may render the battery pack unusable over time. A latter
chapter motivates the need to balance a battery pack and discusses various ways
of cell balancing.

8 Battery Simulator
Based on the DC ECM discussed in this section, the following MATLAB codes
can be used to simulate the voltage and current across a battery. Example 3
shows a use case of the following battery simulator.
function [V, SOC, Vo] = battSIM(I, Batt, Delta)
%% Reading battery ECM parameters
Kbatt = Batt.Kbatt;
Cbatt = Batt.Cbatt ;
K0 = Kbatt(1); K1 = Kbatt(2); K2 = Kbatt(3);
K3 = Kbatt(4); K4 = Kbatt(5); K5 = Kbatt(6);
K6 = Kbatt(7); K7 = Kbatt(8);
R0 = Batt.R0; R1 = Batt.R1; R2 = Batt.R2;
C1 = Batt.C1; C2 = Batt.C2;
ModelID = Batt.ModelID;
E = Batt.SOCsf; % scaling factor
alpha1=exp(-(Delta/(R1*C1)));
alpha2=exp(-(Delta/(R2*C2)));
%% Hysteresis model
h = 0;
%% Determination of SOC (Coulomb counting)
SOC=zeros(length(I),1); SOC(1)=.5;
for k=2:length(I)
SOC(k)= SOC(k-1)+(1/(3600*Cbatt))...
*I(k)*Delta;
if SOC(k) < 0
error('Battery is Empty!!')
elseif SOC(k) > 1
error('Battery is Full!!')
end
end
%% Determination of OCV
zs = (1 - 2*E)*SOC + E;
Vo = K0+K1./zs + K2./(zs.^2) + ...
K3./(zs.^3)+K4./(zs.^4) + ...
K5*zs + K6*log(zs)+ K7*log(1-zs);
%% Determine current through R1 and R2
x1=zeros(length(I),1);
x2=zeros(length(I),1);
for k=1:length(I)
x1(k+1)=alpha1*x1(k)+(1-alpha1)*I(k);

29
 x2(k+1)=alpha2*x2(k)+(1-alpha2)*I(k);
end
i1=zeros(length(I),1); i2=zeros(length(I),1);
for k=1:length(I)
i1(k)=x1(k+1); i2(k)=x2(k+1);
end
%% Determination of terminal voltage
V=zeros(length(I),1);
switch ModelID
case 1
V= I*R0;
case 2
V= I*R0+Vo+h;
case 3
V= I*R0+i1*R1+Vo+h;
case 4
V= I*R0+i1*R1+i2*R2+Vo+h;
end
end

9 Summary
The goal of equivalent circuit model representation of a battery is to have the
ability to represent important states of the battery, such as the state of charge,
available power, and state of health, in terms of non-invasively and continuously
measurable quantities. The electrical equivalent circuit model (ECM) consists
of the following elements:

• Open circuit voltage model. The open circuit voltage of a battery has a
monotonous and non-linear relationship with the state of charge — as the
state of charge increases, the OCV increases and vice versa. The OCV-
SOC relationship is unique to each battery chemistry, size, and physical
shape. The non-linear OCV-SOC relationship is required for battery man-
agement functionalities such as SOC estimation. For a given battery cell,
the parameters of the OCV-SOC relationship is estimated in laborato-
ries. Detailed explanation of OCV parameter estimation (using the least
squares method) is reserved for a separate chapter.
• Series resistance. The series resistance of a battery changes with temper-
ature and age; it also serves as a measure of battery age. Hence, accurate
estimation of the series resistance is important for effective battery man-
agement.
• Relaxation effect. The relaxation effect is due to the response of a battery
due to that of a capacitance along with the internal resistance.

30
 • Hysteresis effect. The memory effect of the battery response is defined as
hysteresis. Hysteresis modelling and its parameter identification remains
a challenging research problem. Latter chapters will employ several tech-
niques to account for the effect of hysteresis when estimating other battery
parameters.

References
[1] M. S. Ahmed, S. A. Raihan, and B. Balasingam, “A scaling approach for
improved state of charge representation in rechargeable batteries,” Applied
Energy, vol. 267, p. 114880, 2020.

[2] G. L. Plett, Battery management systems, Volume I: Battery modeling.

Artech House, 2015.

31