Balancing and SOC Estimation in a Battery

Management System for Electric Vehicle

2021 44th International Spring Seminar on Electronics Technology (ISSE) | 978-1-6654-1477-7/20/$31.00 ©2021 IEEE | DOI: 10.1109/ISSE51996.2021.9467557

Ana-Maria Petri1), Dorin Petreuș1)

1)
Applied Electronics, Technical University of Cluj-Napoca, Cluj-Napoca, Romania
[email protected]

Abstract: This paper presents the implementation of a passive and an active cell balancing method
using MATLAB/Simulink simulations. For passive balancing two variants were implemented: one using
the cell voltage to determine whether balancing is needed, and the second one using the cell SOC. The
SOC was estimated with Coulomb Counting and Extended Kalman Filter, comparing the two algorithm
results, Extended Kalman Filter has smaller estimation error. The results show that the variant based
on the SOC has higher performance. Active balancing was implemented using switched capacitors
topology and energy is transferred from higher charged cells to lower charge ones, with higher efficiency
than passive balancing

1. INTRODUCTION The amount of energy stored in a cell is known as

the state of charge (SOC), and it cannot be directly
Nowadays, electric vehicles (EVs) are already the measured, it can only be estimated. The accuracy of
accepted solution to reduce pollution and consumption SOC estimation is an important problem in EV
of fossil fuel. The development of electric vehicles was application and several methods have been developed
possible based on the advances done for energy storage in the last decade [5].
systems. The most important part of an electric vehicle
is the energy storage system and the main component In this paper, passive and active balancing methods
of this system is the battery [1]. and SOC estimation algorithms for battery packs are
analyzed. The paper is organized as follows: in Section
For an electric vehicle, the battery is the main 2 is presented the importance of balancing and
component, it has the highest cost, biggest weight and classification of balancing methods. Section 3 is
volume in the vehicle [2]. Lithium-ion batteries proved devoted to SOC estimation algorithms and Section 4
to be the best solution for energy storage element in EV illustrates the implementation and the obtained results.
applications, because they have high energy density and Based on the results, some conclusions are given in
high specific power. But, the disadvantage of Li-ion Section 5.
batteries is that they need an energy management
system to assure the monitoring and protection of the 2. BALANCING
battery, to assure operation in the safe area. Beside
The differences between the cell of a battery pack
protection, for battery packs with many cells, the
can be from manufacturing process, such as self-
energy management system must provide cell
discharge rate, storage capacity, aging coefficient, or
balancing, to prolong the life of the battery pack [1, 3].
within the pack the cells can have different operating
The power required by an EV application is higher temperature, therefore in time, during charge-discharge
than the capabilities of a battery cell, so an EV battery, cycles there will be differences between the parameters
known as a battery pack, contains many battery cells, of the cells. During charging some cells can reach full
which are connected in series, to obtain high voltage, charge state quicker than others, or during discharge
and in parallel, for big capacity. The balancing ensures some cells can be deep discharged. The cells in a pack
that every cell is charged and discharged with the same without balancing can face one of these four abusive
amount of energy, and the cells are not abused during situations: overcharge, undercharge, over discharge and
charging or discharging [4, 5, 6]. under discharge [4, 5].

978-1-6654-1477-7/21/$31.00 ©2021 IEEE 1 44th Int. Spring Seminar on Electronics Technology (ISSE)

Authorized licensed use limited to: KLE Technological University. Downloaded on January 15,2022 at 16:36:26 UTC from IEEE Xplore. Restrictions apply.
 All the four mentioned issues are affecting the cell - Adjacent Cell-to-Cell transfer (AC2C): energy is
capacity and health, and eventually the cell will fail. transferred to and from adjacent cells [4].
The failure of one cell led to a fall of the entire pack.
- Directed Cell-to-Cell (DC2C): energy is transferred
So, a balancing method is required for optimal
directly from the highest charged cell to the lowest
operation of the battery pack and longer life.
charged one [4].
There are two main types of balancing: passive and
- Cell-to-Pack (C2P): energy is transferred from the
active balancing.
highest charged cell to pack [4].
In passive balancing the excess energy is dissipated,
- Pack-to-Cell (P2C): energy is transferred from every
from overcharged cells. The dissipation can be done in
cell in the pack to the lowest charged cell [4].
two manners: with a fixed dissipative resistor
connected in parallel with the cell, energy will be - Multi-Cell-to-Multi-Cell (MC2MC): energy is
dissipated all the time through the resistor. In the transferred from the highest charged cells to lowest
second manner, electronic switches are required, and charged cell. MC2MC is like DC2C, but it works
the resistor is connected in parallel with the cell only with more cells [4].
when the cell is overcharged. For switching commands Active balancing methods presented before can be
and deciding which cell is overcharged a control loop implemented in several topologies, which can be
and monitoring the cells voltage are required [4, 5, 7]. classified into three types: capacitor based, inductor
Passive balancing is simple to implement but is less based, and transformer based [4]. The operation of each
efficient because, the energy is dissipated as heat and active balancing topology is based on one of the transfer
can cause a temperature rise in the pack. Also, the methods presented before. In [4] is presented a detailed
capacity of the whole pack is reduced, because energy classification of active balancing topologies. Figure 2
is wasted as heat [4, 5, 7]. In figure 1 is presented the illustrates an active balancing topology with switched
topology of passive balancing using switched resistors: capacitors, which it works using AC2C energy transfer
method.

Fig. 1. Passive balancing with switched resistors Fig. 2. Active balancing with switched capacitors
In active balancing, energy is transferred from the Beside the method and topology used for balancing,
overcharged cells to the less charged ones. This method an important part of the balancing strategy is the
is more efficient than passive balancing and the battery parameter based on with is decided if the balancing is
pack capacity is not affected [4, 5, 7]. needed. Also, some balancing methods need a way to
determine the highest and the lowest charged cells.
The energy transfer is done using capacitive or
Widely used are balancing strategies using the cell
inductive storage component, so the energy is voltage, it requires only voltage sensing and then the
temporarily stored from an overcharged cell and then control used the voltage values to generate the
transferred to an undercharged cell [4, 5, 7]. appropriate balancing commands. Another parameter
The active balancing method can be classified based used for balancing is the SOC of each cell. This strategy
on the way the energy is transferred [4]: needs sensing of the cell voltage or current, an SOC

2 44th Int. Spring Seminar on Electronics Technology (ISSE)

Authorized licensed use limited to: KLE Technological University. Downloaded on January 15,2022 at 16:36:26 UTC from IEEE Xplore. Restrictions apply.
estimation algorithm and based on the SOC values the temperature or number of charge/discharge cycles.
balancing commands are generated [5, 6]. Beside the parameters mentioned before, EKF consider
the noise from the measured data and eliminated it [10].
SOC estimation is another widely discussed topic
when it comes about batteries. Several SOC estimation 4. IMPLEMENTATION AND RESULTS
algorithms exist. Section 3 presents an overview of the
SOC estimation algorithms. The implementation of a battery management
system used to analyze balancing methods and SOC
3. STATE OF CHARGE ESTIMATION estimation algorithms was done in MATLAB/Simulink
environment. The system has two main components:
The state of charge of a battery is defined as the
battery management system (BMS) and battery pack.
ration between the actual charge stored in the battery
From the battery pack information about the voltage,
and the total charge that can be stored in the battery.
current and temperature are sent to the BMS. Based on
The SOC is affected by several factors as capacity, the information received from the battery pack, the
temperature, age, internal impedance, self-discharge BMS process the information and generates balancing
rate, material degradation, electrochemical reactions commands, which are sent back to the battery, compute
and other cell characteristics.[8][9] the SOC, and, in case of faults, safety signals, to prevent
Several SOC estimation methods were developed damages. In figure 3 is presented the block diagram of
and there are five categories as follows: coulomb the simulated system:
counting method, open-circuit voltage method, model-
based method, machine learning method. Coulomb
Counting and open-circuit voltage methods are simpler,
but they have low efficiency. Model-based and machine
learning required bigger computation resources, but
they assure a higher efficiency. [8][9]
In Coulomb Counting method the SOC is estimated
based on the initial SOC of the cell, the capacity of the
cell, and by integrating in time the charge and discharge
current of the cell, using the coulombic efficiency (η)
of the cell. Equation (1) describes this estimation Fig. 3. Block diagram of the simulated system.
algorithm:[10]
The SOC estimation block computes the SOC of the
1 whole battery pack and the SOC of every cell, based on
𝑆𝑂𝐶 𝑡 𝑆𝑂𝐶 ⋅ 𝜂𝐼d𝑡 (1)
𝐶 the voltages of each cell and the current. In figure 4 is
Kalman filters are part of model-base category and presented a comparison of the SOC of a cell using a
they are successfully used for SOC estimation. Using Coulomb Counting, Extended Kalman Filter and the
Kalman filter means that the current state of charge is real SOC of the cell, given by the Simulink cell model,
estimated based on the previous estimation and the during discharging, charging and balancing states.
actual measurements. A Kalman filter can be From figure 4, the SOC estimated with EKF is
implemented in several variants, but for SOC almost identically with the real one, while Coulomb
estimation, the widely used one is the Extended Kalman Counting has an error in estimating the SOC. So, for the
filter (EKF), a nonlinear Kalman filter [9]. balancing methods that use a balancing parameter the
For the implementation of a EKF for SOC online SOC based on the EKF will be used for high efficiency.
estimation, some parameters are needed as follows: The battery pack used in the simulation has six cells
battery capacity, the initial value of the SOC, battery connected in series. To have an imbalance in the pack,
charge/discharge current, battery model, a way to the capacity of four cells was set to the same value, and
predict the SOC based on the actual measurements, and for the other two, cell 4 has a little higher capacity, and
the previous SOC estimated by the filter [9]. For high cell 3 has a little smaller capacity. So, for an analysis of
efficiency estimation can be used some additional the balancing methods a passive balancing method and
parameters, which are affecting the SOC, like an active balancing one were implemented.

3 44th Int. Spring Seminar on Electronics Technology (ISSE)

Authorized licensed use limited to: KLE Technological University. Downloaded on January 15,2022 at 16:36:26 UTC from IEEE Xplore. Restrictions apply.
 Fig. 4. Block diagram of the simulated system.
Passive balancing was implemented based on the Fig. 6. Cell SOCs for voltage based passive balancing
switched resistors topology from figure 1. For the first The passive balancing system was tested using as
variant, cell voltage was used as the balancing the balancing parameter the SOC of the cells, estimated
parameter. In figure 5 is presented the cell voltages, in with EKF. In figure 7 and 8 the cells voltages and SOCs
case of passive balancing with cell voltage as balancing are presented.
parameter:

Fig. 7. Cell voltages for SOC based passive balancing

Fig. 5. Cell voltages for voltage based passive balancing
In the same manner, as in the voltage based passive
Also, the cells SOC estimated with EKF are balancing, the voltages and the SOCs of the cells
illustrated in figure 6. It can be seen from both figures become almost equal, with the dissipation of the energy
5 and 6 that at a certain point, in time, the cell voltages excess from the highest charge cells. But, in this case,
and, also, the cell SOCs, become almost equal, the the goal is to have almost equal SOCs, not voltages. So,
slight difference is small or equal to the value of the in figure 8, the differences between the SOCs
voltage threshold. This equalization is done by are negligible, and the voltages, from figure 7, reach an
dissipating the excess energy from the highest charged equilibrium state. SOC based passive balancing has
cells, until they equal the lowest voltage, which is the higher efficiency because is more important to have the
voltage of cell 3, the one with the lowest capacity in the same SOC in all cells, the voltage can be influenced by
pack, by dissipation. temperature and even the actual SOC of the cell.

4 44th Int. Spring Seminar on Electronics Technology (ISSE)

Authorized licensed use limited to: KLE Technological University. Downloaded on January 15,2022 at 16:36:26 UTC from IEEE Xplore. Restrictions apply.
 Fig. 8. Cell SOCs for SOC based passive balancing Fig. 10. Cell SOCs for active balancing
The passive balancing seems to fulfill the task of In active balancing, as it was mentioned in section 2,
reaching the same SOC for the cells in the pack, but excess energy is transferred from the highest charged cell to
with the cost of dissipating energy, so, energy is loss as the lowest charged one. In figure 9 and 10, cells voltages and
heat. The over all energy stored in the pack is reduced SOCs have an equilibrium value somewhere between the
highest charged cell and the lowest charged one. For active
and the pack will discharge quicker. For high power
balancing the energy is not lost by dissipation, it is
systems, the active balancing method is recommended. transferred. Active balancing improves the capacity of the
A switch capacitor topology, as in figure 2, was battery pack, energy is not lost by dissipation, so, it has a
implemented based on the same system, from figure 3. higher efficiency than passive balancing.
For this topology the balancing is done on the basis of
a PWM signal and the energy is transferred between 5. CONCLUSIONS
two adjacent cells. In figure 9 the cells voltages for
active balancing are presented: In this paper were discussed the topics of cell
balancing and SOC estimation for a battery pack. Using
MATLAB/Simulink passive and active balancing
methods were implemented. Passive balancing was
implemented in two variants: one using cell voltage as
the parameter to determine whether balancing is
needed, and the other using cell SOC. For SOC
estimation, Coulomb Counting and Extended Kalman
Filter were implemented. The results show that
Extended Kalman Filter has smaller estimation error. A
switched capacitor, active balancing method, was used,
resulting in a higher efficiency than in passive
balancing.

REFERENCES
[1] M.A. Hannan, M.M. Hoque, A. Mohamed, A. Ayob,
Review of energy storage systems for electric vehicle
applications: Issues and challenges, Renewable and
Sustainable Energy Reviews, Volume 69, 2017, Pages
Fig. 9. Cell Voltages for active balancing 771-789

5 44th Int. Spring Seminar on Electronics Technology (ISSE)

Authorized licensed use limited to: KLE Technological University. Downloaded on January 15,2022 at 16:36:26 UTC from IEEE Xplore. Restrictions apply.
[2] M. Ehsani, Y. Gao, S. E. Gay, A. Emadi, “Modern [7] D. V. Cadar, D. M. Petreus and T. M. Patarau, "An
Electric, Hybrid Electric, and Fuel Cell Vehicle”, CRC energy converter method for battery cell balancing,"
PRESS, USA, 2000 33rd International Spring Seminar on Electronics
[3] S. Manzetti, F. Mariasiu, “Electric vehicle battery Technology, ISSE 2010, Warsaw, Poland, 2010, pp.
technologies: From present state to future systems”, 290-293
Renewable and Sustainable Energy Reviews, Volume [8] Z. C. Gao, C. S. Chin, W. D. Toh, J. Chiew, J. Jia, "State-
51, 2015 of-Charge Estimation and Active Cell Pack Balancing
[4] J. Carter, Z. Fan, J. Cao, “Cell equalisation circuits: A Design of Lithium Battery Power System for Smart
review”, Journal of Power Sources, Volume 448, 2020 Electric Vehicle", Journal of Advanced Transportation,
vol. 2017, Article ID 6510747, 14 pages, 2017
[5] A. Samanta, S. Chowdhuri, “Active Cell Balancing of
Lithium-ion Battery Pack Using Dual DC-DC Converter [9] Prashant Shrivastava, Tey Kok Soon, Mohd Yamani
and Auxiliary Lead-acid Battery”, Journal of Energy Idna Bin Idris, Saad Mekhilef, Overview of model-based
Storage,Volume 33, 2021 online state-of-charge estimation using Kalman filter
family for lithium-ion batteries, Renewable and
[6] T. Wu, F. Ji, L.Liao, C. Chang, “Voltage-SOC balancing Sustainable Energy Reviews, Volume 113, 2019
control scheme for series-connected lithium-ion battery
packs”, Journal of Energy Storage, Volume 25, 2019

6 44th Int. Spring Seminar on Electronics Technology (ISSE)

Authorized licensed use limited to: KLE Technological University. Downloaded on January 15,2022 at 16:36:26 UTC from IEEE Xplore. Restrictions apply.