IEEE Final only

‑
Abstract—The paper provides a comprehensive discussion of
the necessary considerations in the design of an embedded
system for measuring and estimating the State-Of-Charge
(SOC) of an electric vehicle battery pack. Lithium-ion battery
characteristics are described and used to form an estimation
method, by identifying the relevant external characteristics.
Voltage, current and temperature sensing is discussed as well as
an overview of some current estimation algorithms. Finally
practical hardware and software issues are presented with
suggested solutions. This paper is not intended to provide a
specific design, but rather a general discussion, and therefore
the detailed math or algorithms involved are not included.
"Keywords - electric vehicle; State-of-charge; Fuel gauge;
embedded automotive.
I. INTRODUCTION
There is a growing interest in electric vehicles (EV), as
they represent part of a potential solution to a number of
major challenges facing nations around the world. Their
ability to be entirely fuelled by renewable energy sources is
important in lowering carbon emission, pollution and smog.
Volatile petrol prices have also driven interest [1]. Finally
potential performance advantages over petrol vehicles is
another reason for the renewed focus on the development of
electric car technology and the market segment [2].

The fuel gauge of any vehicle is an important aspect, and
should be able to reliably and accurately give an estimation
of the remaining energy capacity. Modern petrol fuel gauges
are quite sophisticated, accounting for movement, slope and
temperature of the fuel within the tank [3]. Estimations of
travel distance and real-time fuel consumption have also
been implemented.

A different set of challenges exist in providing these
same services in an EV. Firstly, unlike a petrol tank, where
the level of fuel can be directly measured, there is no way
to directly measure a battery’s state-of-charge (SOC) [4].
Instead it’s internal state must be modelled by use of it’s
external input and output. Safety concerns are also
significantly different, as large battery packs rated at
hundreds of volts are required for high torque applications
[5].
II. SENSING AND ESTIMATION
A. Battery Characteristics
Lithium-ion and Lithium-polymer batteries, due to their
high energy to weight ratio, have become a widely used
battery technology across many applications, and in electric
vehicles have become the singular choice of energy storage
at present[6]. The main advantages of lithium-ion batteries
are [7]:
1) High single cell voltage (3-4V);
2) High specific energy;
3) High energy efficiency;
4) Long life.
"
Additionally, the mass production processes already in
practice for lithium-ion cells, destined for consumer devices,
prove the ability to manufacture these cells in high volumes
which will be required for the EV market. All these reasons
set lithium-ion as the best available technology for the
modern electric vehicle.
"
However, batteries are highly complex electro-chemical
systems with non-linear characteristics, thereby proving non-
trivial to model for state-of-charge estimation purposes [6].
There are a great number of factors that will impact the
performance of a lithium-ion battery, such as temperature,
age, charge history and rate of discharge [8]. Figure 1.
shows the general battery characteristic with increasing
temperature to illustrate this non-linear behaviour.
"
Apart from the decreasing performance with discharge at
lower temperatures, a decrease in performance also occurs
with higher discharge rates.
Figure 1.
"
This complexity must be accounted for in the estimation
process if accurate results are to be obtained, which is
important furthermore if high efficiencies, correct charging
R.Chase, RMIT, IEEE

methods, safety and reliability are to be achieved in the
overall system, which depends on the SOC for it’s operation.
"
B. Sensors and Measurements
The internal state of a battery, unlike a conventional petrol
vehicle tank cannot be directly measured by any practical
means. As such, the external characteristics of the battery
must be monitored, and by using an algorithm, the internal
state can be estimated. Following is a discussion of the three
main measurements used in calculating SOC. Figure 2.
illustrates a system-level diagram of an example SOC
estimation device.
1) Voltage
The open circuit voltage of the cell
can be an accurate reflection of a the
SOC, particularly near the empty and full states of the cell,
however this method requires that the battery is not in use,
and has not been for a period of time [9]. Within electric
vehicles, this is not a practical method, as the SOC is most
important to the driver exactly when the vehicle is in motion.
Hence the in-circuit voltage measurement across the battery
must be used.
However, it must be noted that vehicle battery voltages
can be excessively high and therefore that a high quality
method of isolation is required, as well as some method of
stepping down the voltage to an appropriate level,
measurable by embedded electronics. The basic ‘resistor
divider network’ would be the most simple way to achieve
the desired voltage, however very high resistance values are
required to minimise current through the network.
Further efficiency gains can be achieved by implementing
a switching system into the network, such as a series of
switching transistors, and only allowing current to flow
through at the sampling moment. When implementing this
method however, care should be taken to ensure that the
sample is taken a sufficient period after the switches have
been turned on to allow the network to achieve a stable and
reflective voltage.
The usual maximum voltage of a single charged lithium-
ion cell is approximately 4.2V, and at empty point, drops no
lower than 3.5V. The individual cell voltages sit well within
the typical limits of a microcontroller’s analog-to-digital
converter peripherals, which provides a simple and
appropriate division ratio to be implemented in the resistive
divider network, namely the ratio should equal the number
of cells in series in the battery pack.
"2) Current
An ampere count can provide a simple and reliable
method of estimation, and is applicable to all battery types
[10]. Efficiencies of conversion need to be accounted for in
the model, and the current measurements must be accurate.
Two basic methods of current measurement widely
employed are measuring the voltage across a sense resistor,
or the use of a hall-
effect sensor.
The advantages of a
hall effect sensor is
the effective isolation
a n d s i m p l i c i t y,
e s p e c i a l l y w h e n
employed as a voltage
output device. It
negates the need for
current to voltage
conversion, which is
performed as a by-
product of the method
o f s e n s i n g .
S e n s i t i v i t y d r i f t
however is major
issue, as temperature
variations can have a dramatic impact on the performance of
the sensor [11]. Mechanical reliability may also be an issue
to consider with such devices.
The second method, whereby the current is fed through a
resistor, and the voltage measured, alleviates some of the
issues of a hall-effect sensor. For instance, while temperature
has an effect on the system, it does not effect the sensitivity
of the system, but rather the resistance is increased, simply
causing the measured voltage to be higher. Such an effect
can be easily accounted for in software along with a
temperature measurement on the resistor, especially
considering how well known and studied resistance with
respect to temperature is.
When considering energy losses however, the resistor
must be chosen to be extremely small. Additionally the
voltage output must be within the measurable range of the
voltage sensor, which if, most likely, is performed by an
integrated CMOS circuit, must be less than 5V. Depending
on the expected current output of the battery this leads to a
predefined maximum and minimum resistance value.
"3) Temperature
As mentioned, temperature effects the performance of
both the battery itself, as well as the sensors measuring it.
Therefore it an important aspect in the estimation process if
accuracy is to be achieved. Ideally temperature values
should be made available to the estimation algorithm from
Figure 2.

the battery cells temperature, any resistors used in the
voltage or current measurements, or in the case of a hall-
effect sensor, the temperature of the coil. The effects of
temperature increases should then be accounted for in the
software of the application.
C. Estimation
With the growing interest in electric vehicles, estimation
methods have received much attention and research of late.
Very simple methods exist, currently in use in various
consumer electronics, such as ampere counting and open-
circuit voltage, however the accuracy required for electric
vehicles demands greater care.
Neural network estimation has been put forward as an
intelligent, self-learning method of SOC estimation
achieving a square error less than 1% [12].
Multi-state technic along with a Kalman filter is another
method suggested, which achieves a square error of 2.72%
[13] . While a prediction algorithm based on a least squares
support vector machine claims to achieve greater accuracy
than neural network estimation by the fact that the maximum
error is only 2% compared to 3% for neural network [14].
The advantage of the neural network is that the self-
learning feature supplements large amounts of experimental
data, and continues learning throughout operation. The
model based predictors however require a firm experimental
database in order to fulfil the accuracy claimed.
With the emerging Internet of Things technologies, the
neural network and other self-learning methods may indeed
prove most effective, with the ability to calibrate itself using
data from the entire global vehicle fleet and moving the
processing into the cloud.
III. PRACTICAL HARDWARE AND SOFTWARE CONSIDERATION
A. Electromagnetic Interference (EMI)
Electromagnetic interference produced by the power
supply and motor system within an electric car, presents a
substantially different EMI environment as compared to
petrol vehicles. The typical electric vehicle power system
contains high voltage DC/DC converters and three-phase
induction motors, presenting large switching voltages and
high current flow. Figure 3. shows the measured current
(yellow) and voltage (blue) in the cabling between a DC-DC
switching power supply and electric motor.
Existing research suggests that the EMI within electric
vehicles is likely to fall around 10 MHz for common mode
interference and between 10 kHz 100 kHz for differential
mode interference. Common mode noise is quite minimal in
all the experimental and model evidence, whereas
differential is significant, with peaks up to 2.5 A [15].
The use of CM inductors to filter out the currents appears
most effective. These however are significant in weight and
size. Inclusion in a battery management system, or for that
matter, in any specific embedded system is not ideal.
The suggested compromise is to use a RC filter, which
though less effective is far smaller and lighter.
Use of electrolytic capacitors in the filtering circuitry has
limited effectiveness. It is a known phenomena that
Figure 3.
"capacitors decrease in impedance at higher frequencies to
the resonant point, after which the impedance increases due
to self-inductance [16]. For electrolytic capacitors, this
effect occurs at relatively low frequencies , this makes them
rather unsuited to filtering higher frequencies. For example
an 150µF etched aluminium foil capacitor at 25℃, 20V has
an impedance over 5 Ω at 100 Hz, but decreases to a
minimum just over 0.1 Ω at 1 MHz [16].
"Ceramic capacitors impedance curves also have similar
characteristic due to parasitic inductances, however they
occur at different frequencies than electrolytic capacitors.
For instance a 0.1µF ceramic capacitor has a minimum
impedance at about 10.5 MHz [17]. A low valued ceramic
capacitor can be placed in parallel with the electrolytic then
and the combination of both types of capacitor provides a
wide spectrum filter, where in the example case using 150µF
electrolytic and 0.1µF ceramic capacitors will provide
effective filtering up to 10 MHz.
"In addition to filtering the signal lines, care should be taken
in the layout and routing of the PCB design. Armstrong
recommends three main techniques, being circuit
segregation, interface suppression, ground and power planes
[18]. Circuit segregation is not relevant to the SOC device
design, as the dominant EMI is expected to be from external
sources which will be present across the entire area of the
device. Interface suppression is implemented to an extent as
a byproduct of the isolation between high and low voltage
sections of the circuit. This technique however is more
tailored towards preventing noise from one part of the circuit
infecting another part, which in this case is not the major
concern. Ground and power planes however will be of use
in minimising EMI coupling."
B. Isolation
Given that the vehicle battery voltage may exceed 400V,
isolation is safety requirement. The SOC measuring device
is likely to be connected via cabling to other areas of the
vehicle, for communication or the power supply. In the
event of component failure, these connections cannot

transmit the battery voltage to unexpected areas of the
vehicle.
The isolation circuitry will induce distortion into the
signals travelling either side of the isolation, so the hardware
design should keep analog signals on the battery side, and
transmit a digital signal through the isolation. This allows
for the use of power efficient IC isolation such as digital RF
isolators. Figure 4. shows a digital 1-Wire signal before
(green) and after (yellow) passing through an isolation
circuit using an RF isolator, showing some small distortion
that would affect the accuracy of an analog signal, but is of
no consequence with a digital signal.
Figure 4.
"The final consideration in isolation is to ensure sufficient
clearances between the circuit tracks or wires. IPC-2221
standards suggests an 0.8mm clearance between 450V tracks
for a coated print circuit board [19].
C. Resistors
For accurate measurement results, the tolerances of the
resistors used in the voltage and current measurements need
to be taken into account. The resistive divider network can
be designed with parallel resistors to allow for fine tuning,
but the actual resistance should ideally be measured and
accounted for via software. It should also be noted that there
is likely to be a limit on the resistance values that can be
used, set by the input impedance of the device measuring the
voltage across the divider. As such, the current draw will
have a minimum achievable value for any given division
ratio, and hence a suitable power rating is required for the
chosen resistors.
IV. CONCLUSION
The paper describes the considerations and theory that are
required in the design and construction of a State-of-charge
measuring embedded system specific to electric vehicle
batteries. The measured signals and various estimation
methods were discussed and compared, and some practical
solutions to EMI and isolation requirements were put
forward.
REFERENCES
1. T.Trigg and P.Telleen, “Global EV Outlook,” IEA, Paris, France,
Rep. 1, April 2013
2. Maini, Chetan; Gopal, Kartik; Prakash, R., "Making of an ‘all
reason’ electric Vehicle," Electric Vehicle Symposium and Exhibition
(EVS27), 2013 World , vol., no., pp.1,4, 17-20 Nov. 2013
3. Goodyer, E., "A Microprocessor Controlled Level Gauge For Use In
Petrol Road Tankers," Measurements we Couldn't Make Without a
Micro, IEE Colloquium on , vol., no., pp.1,3, 17 May 1988
4. Jie Xu; Mingyu Gao; Zhiwei He; Jianbin Yao; Hongfeng Xu,
"Design and Study on the State of Charge Estimation for Lithium-ion
Battery Pack in Electric Vehicle," Artificial Intelligence and
Computational Intelligence, 2009. AICI '09. International
Conference on , vol.3, no., pp.316,320, 7-8 Nov. 2009
5. Terras, J.M.; Sousa, D.M.; Roque, A.; Neves, A., "Simulation of a
commercial electric vehicle: Dynamic aspects and performance,"
Power Electronics and Applications (EPE 2011), Proceedings of the
2011-14th European Conference on , vol., no., pp.1,10, Aug. 30
2011-Sept. 1 2011
6. Jun Xu; Mi, C.C.; Binggang Cao; Junjun Deng; Zheng Chen; Siqi Li,
"The State of Charge Estimation of Lithium-Ion Batteries Based on a
Proportional-Integral Observer," Vehicular Technology, IEEE
Transactions on , vol.63, no.4, pp.1614,1621, May 2014
7. Horiba, T., "Lithium-Ion Battery Systems," Proceedings of the
IEEE , vol.102, no.6, pp.939,950, June 2014.
8. MAXIM 3958.
9. Xianmin Li; Yuanlei Yang, "Research on the calculation method and
dynamic modeling of traction-battery SOC for electric vehicle,"
Mechanic Automation and Control Engineering (MACE), 2011
Second International Conference on , vol., no., pp.6778,6782, 15-17
July 2011
10. Wang NianCHun; Qin Yan, "Research on State of Charge Estimation
of Batteries Used in Electric Vehicle," Power and Energy
Engineering Conference (APPEEC), 2011 Asia-Pacific , vol., no., pp.
1,4, 25-28 March 2011
11. Ajbl, Andrea; Pastre, M.; Kayal, M., "A fully integrated Hall sensor
microsystem with current-mode output," Electronics, Circuits and
Systems (ICECS), 2011 18th IEEE International Conference on ,
vol., no., pp.464,467, 11-14 Dec. 2011
12. Zhou Yongqin; Zhang Yanming; Zhao Pengshu; Han Chunli, "Study
of battery state-of-charge estimation for hybrid electric vehicles,"
Strategic Technology (IFOST), 2011 6th International Forum on ,
vol.1, no., pp.287,290, 22-24 Aug. 2011
13. Li Yong; Wang Lifang; Liao Chenglin; Wang Liye; Xu Dongping,
"State-of-Charge Estimation of Lithium-Ion Battery Using Multi-
State Estimate Technic for Electric Vehicle Applications," Vehicle
Power and Propulsion Conference (VPPC), 2013 IEEE , vol., no.,
pp.1,5, 15-18 Oct. 2013
14. Hui Bao; Yang Yu, "State of Charge Estimation for Electric Vehicle
Batteries Based on LS-SVM," Intelligent Human-Machine Systems
and Cybernetics (IHMSC), 2013 5th International Conference on ,
vol.1, no., pp.442,445, 26-27 Aug. 2013
15. Jing Xue; Wang, F., "Modeling and design of common-mode
inductor for conductive EMI noise suppression in DC-fed motor
drive system," Energy Conversion Congress and Exposition (ECCE),
2012 IEEE , vol., no., pp.645,651, 15-20 Sept. 2012
16. Campbell, D.S., "Electrolytic capacitors," Radio and Electronic
Engineer , vol.41, no.1, pp.5,, January 1971 doi: 10.1049/ree.
1971.0003
17. Cain, J., “Parasitic Inductance of Multilayer Ceramic Capactiors”,
AVX Corporation Technical Information, August 2004
18. Armstrong, M. K., "PCB design techniques for lowest-cost EMC
compliance .1," Electronics & Communication Engineering Journal ,
vol.11, no.4, pp.185,194, Aug 1999 doi: 10.1049/ecej:19990402
19. Generic Standard for Print Board Design, IPC-2221A, February
1998

IEEE Final only

More Related Content

What's hot (20)

Viewers also liked (13)

Similar to IEEE Final only (20)

IEEE Final only