Next Energy 8 (2025) 100258

Contents lists available at ScienceDirect

Next Energy
journal homepage: [Link]/journal/next-energy

Short communication

An effective passive cell balancing technique for lithium-ion battery

Sudarshan L. Chavan , Manjusha A. Kanawade, Rahul S. Ankushe
⁎ ]]
]]]]]]
]]

Department of Electrical Engineering, JSPMs Rajarshi Shahu College of Engineering, Pune, India

A R T I C L E I N F O ABS TRACT

Keywords: The increasing demand for clean transportation has propelled research and development in electric vehicles
PI controller (EVs), with a crucial focus on enhancing battery technologies. This paper presents a novel approach to a battery
State-of-charge management system by implementing a passive cell balancing system for lithium-ion battery packs. The pro­
Cell balancing posed system employs a proportional-integral (PI) controller to address voltage imbalances among individual
Lithium-ion battery
cells, aiming to improve battery life and longevity without the need for a complex active control circuit. The
study explores performance evaluation under diverse conditions, considering factors such as system capacity
retention, energy efficiency, and overall reliability. Safety and thermal management considerations play a
crucial role in the implementation, ensuring the longevity and stability of the lithium-ion battery pack. The
primary objective of this research is to extend the operational life of lithium-ion batteries, reduce maintenance
costs associated with battery management, and contribute to sustainable energy solutions. In the presented
study, first, a Simulation model is developed in MATLAB, and the results are verified by implementing a
hardware model.

1. Introduction be very difficult to implement practically. The method used for cell
balancing in Shukla and Patel [2] is quite difficult to understand. No
The rapid growth of electric vehicles (EVs) in recent years has un­ safety other than cell balancing is presented in that research. The
derscored the critical role of battery technology in the advancement of technique presented in Song and Lee [3] gives a simulation circuit for
sustainable transportation. Lithium-ion batteries have emerged as the cell balancing. Validation of results with practically implemented cir­
predominant energy storage solution for EVs due to their high energy cuit results is not provided. The combined active and passive cell bal­
density, long cyclic life, and relatively low self-discharge rates. ancing method presented in Kurpiel et al. [4] is applicable for specific
However, the performance and longevity of lithium-ion battery packs environments only. Generalization of this method is quite difficult. The
are subject to various challenges, including cell voltage imbalances that circuit presented in Pham et al. [5] is quite slow, and the time required
can degrade battery capacity and compromise safety. Battery manage­ for balancing cells is also very high. The hardware in loop technique
ment systems (BMSs) play a pivotal role in monitoring and controlling used in Khalid et al. [6] for commercial passive cell balancing of BMS is
the operation of lithium-ion battery packs to ensure optimal perfor­ effective, but it is costly and needs to be used in special environments
mance and safety. Among the key functions of a BMS, cell balancing is only. The literature review of ANN controller-based passive cell bal­
particularly crucial for mitigating voltage differentials among in­ ancing presented in Kurucan and Özbaltan [7] focuses on applications
dividual cells within a pack. Traditional active balancing techniques of ANN for passive cell balancing. The modified proportional-integral
involve the redistribution of charge between cells using complex cir­ (PI) controller-based cell balancing presented in Aslam [8] is not vali­
cuitry, which can introduce additional components, costs, and points of dated with actual experimental results by practically implementing it.
failure. In this context, passive cell balancing systems have emerged as The state of charge and remaining charge estimation of series-con­
an alternative approach to address voltage imbalances while mini­ nected lithium-ion batteries for cell balancing presented in Chun et al.
mizing complexity and cost. Numbers of techniques are presented in the [9] does not clearly indicate which method is actually used for balan­
literature for passive cell balancing; a review of few is given herewith. cing the cells. The practical implementation of the method used is also
In Karmakar et al. [1], artificial neural network (ANN)-based technique not presented. In Paidi and Gudey [10], an overview of a few active and
is used for passive cell balancing. The complex circuitry used seems to passive cell balancing methods is presented. No method is implemented

⁎
Corresponding author.
E-mail addresses: chavan_sl@[Link] (S.L. Chavan), manjushakanawade@[Link] (M.A. Kanawade), rahulankushe123@[Link] (R.S. Ankushe).

[Link]
Received 29 October 2024; Received in revised form 14 February 2025; Accepted 19 February 2025
2949-821X/© 2025 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC license ([Link]
4.0/).
S.L. Chavan, M.A. Kanawade and R.S. Ankushe Next Energy 8 (2025) 100258

Fig. 1. Battery management system.

Fig. 2. Equalization circuit.

or simulated to represent results. In Panchal et al. [11], the design and complex where as many simple methods are already available in lit­
simulation of an inductor-based active cell balancing circuit are pre­ erature. In Samaddar et al. [19], review of battery cell balancing
sented with very less information about passive cell balancing. Passive methodologies for optimizing battery pack performance in EVs is pre­
cell balancing is neither simulated nor practically implemented. In sented. Again it is not clearly presented which method is superior.
GPECOM [12], analyses of cell balancing techniques in BMS for EV are Fuzzy-PI controlled battery equalization for series-connected lithium-
presented. It is very difficult to guess which method is best for passive ion battery strings is presented in Omariba et al. [20]. Hardware im­
cell balancing because relative merits and demerits of presented plementation of the presented method is not given. In Ling et al. [21], a
methods are not given. A passive cell balancing algorithm for charge review of cell equalization methods for lithium-ion and lithium polymer
equalization of series-connected battery cells is presented in Nath and battery systems is presented. A comparative analysis of the presented
Rajpathak [13]. It is also not clear from the presented methods which methods is not given. The research presented in Moore and Schneider
method is more effective. Practical implementation of the presented [22] introduces 3 advanced algorithms to estimate the SoC: deep neural
method and effectiveness of the presented method compared to other network (DNN), gated recurrent unit, and long short-term memory. The
methods are not given. A comparison of cell balancing methods for presented 3 models are trained and validated using laboratory data
energy storage applications is presented in Di Rienzo et al. [14]. Only from a lithium-ion 18650 battery and simulation data from Matlab/
an overview of various methods is given, which method is superior is Simulink for a LiCoO2 battery cell. This study highlights the effective­
not mentioned anywhere. A review of battery balancing methods is ness of the DNN algorithm in providing a reliable state of charge (SOC)
given in Lee et al. [15]. More focus is given to active cell balancing than estimation under diverse operating conditions, showcasing its potential
passive cell balancing. An algorithm for the charge equalization con­ for enhancing battery management in EV applications. The results ob­
troller of series-connected lithium-ion battery cells in EV applications is tained are very good. The work presented in El Fallah et al. [23] pro­
presented in Cao et al. [16]. The practical implementation of the pre­ poses a robust and efficient combined SOC estimation method, gated
sented method is not highlighted. Design and Implementation of a PI- recurrent unit-adaptive Kalman filter (AKF), which combines the gated
controlled lithium-ion battery charger with protection and monitoring recurrent unit recurrent neural network (GRU-RNN) and the AKF.
is presented in Hoque et al. [17]. In the presented research, more focus Firstly, the GRU-RNN is used to establish a mapping model between the
is given to the protection of the charger instead of cell balancing. In battery-measured variables and SOC in the full temperature range and
Maharjan et al. [18], passive cell balancing of Li-Ion batteries used for to achieve SOC preestimation. Then, an AKF is employed to filter the
automotive applications is presented. The presented method is more output SOC of the GRU-RNN for reducing the fluctuation in

2
S.L. Chavan, M.A. Kanawade and R.S. Ankushe Next Energy 8 (2025) 100258

Fig. 3. Fig.3 Simulation model for cell balancing.

preestimated SOC. At last, the accurate and stable estimated SOC is members are replacing by individual cell, and walking speed is replaced
obtained. By comparing with other methods of the same type, the by cell SOC. The rapid rise in EV use has prompted a demand for battery
proposed method is found to be good in SOC estimation performance systems with high ratings, more efficiency, and longer life. The sub­
and computation efficiency, especially since it has good performance in systems of an entire battery system are management, balancing, and
initial SOC convergence ability. The work presented in Chen et al. [24] protection. In terms of the battery with lifespan, the balancing tech­
uses a DNN, an ANN, and a gated recurrent unit recurrent neural net­ nique is the most crucial of the 3 components because without it, the
work (GRU-RNN) to estimate the SOC from data collected in the la­ voltages of the individual cells will move apart over time. The cell
boratory on a Li-ion battery 18650, and from data obtained under unbalancing may overcharge or overheat cells, which may be prone to
Matlab/Simulink software from a LiCoO2 battery cell. This study also accelerate cell degradation. The cell unbalancing can cause fire or even
presents a comparative study between the SOC measured experimen­ explode as a thermal runway condition takes place due to an increase in
tally and the SOC obtained by simulation at different temperatures and cell voltages. The presented work describes how a battery system with a
under various profiles so that the battery could be exposed to variable PI controller uses the proposed technique to control a voltage level of
dynamics. The results obtained are really appreciable. A comprehensive individual cell within battery pack and guarantee that all cells are
review of SOC estimation methods for lithium-ion (Li-Ion) batteries, charged and discharged equally. The presented research actually pro­
with a specific focus on EVs, is presented in El Fallah et al. [25]. Various poses a novel passive cell balancing system for lithium-ion battery
approaches, including data-driven techniques, advanced filtering packs, leveraging a PI controller to regulate voltage differentials among
methods, and machine learning algorithms have been explored to en­ the cells. The presented method is first simulated in MATLAB and then
hance SOC estimation accuracy. In Demirci et al. [26],different state practically implemented to validate the results. The use of a PI con­
estimation methods are provided in MATLAB model. In [Link] et al. troller offers a robust and flexible control strategy, allowing for precise
[27] the robust full-order proportional-integral observer, designed to adjustment of balancing currents while adapting to varying load and
overcome SOC estimation misalignment challenges in EVs during high environmental conditions. The presented combination of theoretical
load shocks and temperature fluctuations. Extensive comparison ex­ analysis, simulation studies, and practical experimentation of passive
periments, in conjunction with a variety of driving data, validate the cell balancing is very useful in real-world EV applications.
robust full-order proportional-integral observer method. The results
indicate that the SOC estimation error remains below 2.11%, even 2. Methodology
amidst unknown model parameters and high-frequency noise inter­
ference. A typical BMS is shown in Fig. 1. Passive cell balancing is a tech­
The presented research actually proposes a novel passive cell bal­ nique used in BMS to equalize the charge among individual cells within
ancing system for lithium-ion battery packs. It is the process of ramping a battery pack without dissipating excess energy as heat [21]. Em­
down the SOC of the cells to the lowest SOC of the cell, which is present ploying a PI controller in passive cell balancing helps to regulate the
in the group or pack. In simple words, consider a family having 5 energy transfer between cells and ensure that they reach a balanced
members, such as parents and children’s. The speed of waking of the state. Initially, the BMS monitors the voltage across each cell in the
parents is faster than the children’s. When the family going to walk for battery pack. Variations in voltage levels among cells indicate an im­
market, all the family members trying to match the speed of slowest balance in their charges.
person in the family, because it is necessary to avoid accidents. It is just The PI controller calculates the error by comparing the SOC levels of
analog to passive cell balancing. Family is replacing by battery pack, each cell to a desired reference SOC or the average SOC of all cells. The

3
S.L. Chavan, M.A. Kanawade and R.S. Ankushe Next Energy 8 (2025) 100258

Fig. 4. System flowchart.

error represents the deviation from the desired balanced state. the cells remain balanced under varying load conditions. The presented
Within the scope of the presented research, an exploration of the method allows the BMS to maintain cell balance efficiently and prevent
intricate dynamics of BMS is undertaken, with a particular emphasis on overcharging or discharging of specific cells, which can lead to reduced
the indispensable role played by the PI controller. The proportional battery life or safety hazards.
component of the PI controller generates an output proportional to the The fundamental equation for cell balancing involves Ohm’s law:
instantaneous error between each cell's SOC and the reference SOC.
V = I × R,
This output determines the amount of energy to be transferred between
cells for balancing. The Integral component of the PI controller con­ where V is the voltage across the resistor, I is the current through the
siders the accumulated error over time. It ensures that any steady-state resistor, and R is the resistance. The energy dissipated by the resistor
error, which might remain due to differences in cell capacities or var­ over time can be calculated using
iations, is gradually corrected. The integral term helps in fine-tuning
the balancing process. The controller then activates passive balancing P = I² × R,
circuits to redistribute energy among cells. For instance, energy from where P is the power dissipated. The balancing time t can be estimated
higher SOC cells could be diverted to lower SOC cells until a balanced using
state is achieved. The PI controller continuously adjusts the energy
transfer rate based on the error signal. As cells reach closer to the ba­ t = Q_imbalance/I,
lanced state, the energy transfer rate decreases to prevent over­
where Q_imbalance is the charge imbalance that needs to be equalized.
correction and oscillation around the set point. Once the cells reach to a
Critical considerations are that it is required to ensure that the balan­
state where their SOC levels are within an acceptable tolerance range of
cing does not overheat the cells. The balancing process must be efficient
each other, the PI controller ensures that the energy transfer ceases and

4
S.L. Chavan, M.A. Kanawade and R.S. Ankushe Next Energy 8 (2025) 100258

Fig. 5. Circuit diagram for hardware implementation.

Fig. 6. Basic block diagram of passive cell balancing.

Fig. 7. Control technique for pulse generation.

5
S.L. Chavan, M.A. Kanawade and R.S. Ankushe Next Energy 8 (2025) 100258

Table 1 associated with control and switching signals. In practical applications

Results of the first run many factors can be taken into consideration such as temperature, wear
Cell No. Initial SOC Final SOC
and tear, improper connections etc. In order to work with higher rating
systems it is necessary to provide proper cooling system to improve
1. 100 79.53 efficiency.
2. 80 79.99 The equalization circuit is shown in Fig. 2. The equalization circuit
3. 90 79.55
in a BMS is responsible for ensuring that all cells reach the same voltage
SOC= state of charge. level. This circuit can be modeled in the topology to understand its
behavior under different operating conditions. It is essential to re­
Table 2 present this circuit accurately in simulations to predict the performance
Results of the second run and identify potential issues. The simulation model developed in MA­
TLAB is shown in Fig. 3, whereas the flowchart of the system is shown
Cell No. Initial SOC Final SOC
in Fig. 4.
1. 83.94 75.69
2. 94.68 75.72 3.2. Hardware implementation
3. 75.73 75.70

SOC = state of charge. The hardware implementation of passive cell balancing using a PI
controller involves several components and circuits to monitor, process,
Table 3 and regulate the energy transfer between cells within a battery pack.
Results of the third run Each cell's voltage and SOC are measured using precision analog front-
end circuits. The circuit involves voltage dividers to scale down high
Cell No. Initial SOC Final SOC
voltages for safe measurement and analog-to-digital converters (ADCs)
1. 60.65 55.49 to digitize the voltage readings. An Arduino UNO microcontroller board
2. 70 55.48 based on the ATmega328P is used to control the developed circuit. A
3. 55.50 55.46 dedicated processing unit executes the PI control algorithm. It reads the
SOC = state of charge. voltage and SOC from the cells, performs calculations, and generates
control signals for balancing. The microcontroller carries out mathe­
enough to prevent significant energy loss. matical computations for both the proportional and integral compo­
The specific implementation details, circuitry, and parameters of the nents of the PI controller. Based on the error calculated from SOC dif­
PI controller will vary depending on the battery chemistry, cell con­ ferentials, the microcontroller generates control signals that regulate
figuration, and balancing strategy employed in the system. the energy transfer between cells. The balancing circuit consists of re­
sistors arranged in circuits to allow controlled energy transfer between
cells. The microcontroller activates these elements based on the control
3. Model Development signals. IGBT-based relays are used as switches to control energy flow
through the balancing circuits. The microcontroller controls them to
3.1. Simulation model development manage energy transfer as directed by the PI controller. For this task, an
operating voltage of 5VDC, a relay current rating of 10 A at 250VAC,
A feedback control system known as a proportional-integral deriva­ and 10 A at 30VDC (noninductive) is used. The microcontroller con­
tive controller is frequently used in industrial control applications to tinuously monitors cell voltages and SOC, executes balancing algo­
control a process variable. The controller operates by monitoring the rithms, and provides feedback to external systems or display units for
difference between the target set point and the actual process variable monitoring purposes. Fig. 5 shows the circuit diagram for the hardware-
continually and modifying a control variable to reduce this difference. In implemented.
the presented system, PI controller takes differences of SOCs of cell as a Fig. 6 shows a basic block diagram, which shows that the output of
reference input signal, and as SOC of cells starts getting dissipate the the PI controller is given to the battery pack through bus selector SOC is
difference will tend to 0. Once the difference becomes 0 PI controller measured, and then it is fed to the input. Reference SOC is determined
stops working. The output of the PI controller is used to control the in­ from the battery pack, and the error is given to the PI controller for
sulated gate bipolar transistor (IGBT) switch The developed simulation further control action. A detailed block diagram of pulse generation is
model consist of PI controller, constant block, IGBT switch, adder, dif­ shown in Fig. 7 where sinusoidal pulse width modulation technique is
ference block, matrix multiplication block, display, resistor, and lithium- used for pulse generation. The generated pulse is given to MOSFET, and
ion cells of 3.6 V, 12 Ah capacity. Three lithium-ion cells having initial energy is dissipated till the reference SOC is achieved.
SOC of 100, 80, and 90, respectively, are first considered for cell bal­ PI controller parameters are given as
ancing. As the minimum SOC from all 3 cells is 80, the minimum
SOC(80) is taken as a reference. At every cycle, the actual SOC is com­ Kp = 1
pared with the reference SOC, and the error is given to the PI controller.
Ki = 0.5
From the output of the PI controller, a gate pulse is generated, the gate
pulse is given to the switch, and cell balancing is achieved. The hardware design and implementation of passive cell balancing
Simulation results and observations show that some cells are not using a PI controller involves a structured approach for monitoring,
dissipated to the desired reference value; there are some minimal losses processing, and dissipating excess energy from overcharged cell voltage

Table 4
Circuit condition before cell balancing

Cell no. Voltage (v) State of charge (%SOC) Error (%) Operation of relay

1. 3.36 83.94% 2.70 ON

2. 3.80 94.68% 3.30 ON
3. 3.03 75.73% 0.00 OFF

6
S.L. Chavan, M.A. Kanawade and R.S. Ankushe Next Energy 8 (2025) 100258

Fig. 8. Circuit before cell balancing.

Fig. 9. Circuit after cell balancing.

Fig. 10. Results before cell balancing.

• Voltage and SOC monitoring system

o The voltage and SOC of each individual cell are monitored using
precision analog circuitry. High cell voltages are scaled down using
voltage divider networks, ensuring safe levels for measurement by
downstream components. These scaled voltages are digitized via
ADCs integrated into the circuit. The ADC outputs provide the mi­
crocontroller with reliable digital representations of the cell vol­
tages, enabling accurate processing and control actions.
Fig. 11. Results after cell balancing.
• Microcontroller unit and algorithm processing
and maintaining a consistent SOC across the battery, thereby enhancing o The heart of the system is the Arduino UNO microcontroller, which
its operational efficiency and lifespan. The process relies on converting utilizes the ATmega328P processor. The microcontroller executes
excess energy from higher SOC cells into heat through a controlled the PI control algorithm, a cornerstone of the balancing strategy.
dissipation mechanism. o PI controller execution:

7
S.L. Chavan, M.A. Kanawade and R.S. Ankushe Next Energy 8 (2025) 100258

Table 5
Circuit condition after cell balancing

Cell no. Voltage (v) State of charge (%SOC) Error (%) Operation of relay
1. 3.03 75.73% 0.00 OFF
2. 3.03 75.73% 0.00 OFF
3. 3.03 75.73% 0.00 OFF

Table 8
Results of the second run with and without PI controller

Cell No. Initial SOC Final SOC Balancing time (sec)

PI controller Without PI
controller

1 83.94 75.69 231 308

2 94.68 75.72 533 711
3 75.73 75.70 000 000

PI = proportional-integral; SOC = state of charge.

Table 9
Results of the second run with and without PI controller

Cell No. Initial SOC Final SOC Balancing time (sec)

PI controller Without PI
controller

1 60.65 55.49 144.75 193

2 70 55.48 408 544
3 55.50 55.46 000 000

PI = proportional-integral; SOC = state of charge.

due to their reliability and suitability in handling the balancing loads,

Fig. 12. Variation of state of charge (SOC) for cell balancing. with the following operational specifications:
Operating voltage: 5 VDC
Table 6 Current rating: 10 A at 250 VAC
Comparison between simulation and hardware results 10 A at 30 VDC (noninductive)
Cell No. Initial SOC Final SOC
• Circuit design and implementation
By simulation By hardware
o A detailed circuit schematic (as illustrated in Fig. 5.) represents the
1 83.94 75.69 75.73
configuration of components, including the microcontroller, ADCs,
2 94.68 75.72 75.73
3 75.73 75.70 75.73 voltage dividers, resistors, and IGBT relays. The diagram demon­
strates the interconnected architecture and highlights the flow of
SOC = state of charge. control signals and energy dissipation paths, emphasizing the sys­
tem's modular and scalable design.
Table 7
Results of the first run with and without PI controller 4. Results and Discussion
Cell No. Initial SOC Final SOC Balancing time (sec)
For the development of the simulation model, 3 lithium-ion cells
PI controller Without PI having capacities 3.6 V and 12 Ah capacities are considered. Three
controller
different sets of readings are taken to validate the model. The results of
1 100 79.53 600 750 3 different runs of the simulation model are shown in Tables 1–3.
2 80 79.99 000 000 The simulation result clearly shows that the developed simulation
3 90 79.55 281 375 model works properly.
Within the context of the presented research on BMSs, the hardware
PI = proportional-integral; SOC = state of charge.
results section unveils crucial insights into the performance of in­
dividual cells within the battery pack. Initially, distinct voltage levels,
Based on the computed error, the proportional and integral com­ SOC, and switch operations underscored the heterogeneous nature of
ponents of the PI controller are calculated. These outputs form the cells. Recognizing the imperative of achieving a balanced charge
control signals that guide the activation of passive balancing ele­ distribution, the presented study implemented a comprehensive cell
ments. balancing procedure. The ensuing data provide a snapshot of the
o Switching and control mechanism transformed state postbalancing, demonstrating a harmonized equili­
brium achieved through meticulous energy management. These results
IGBT-based relays are employed as switches to engage or disengage serve as a cornerstone for a detailed analysis, offering valuable insights
the resistive balancing loads. The microcontroller controls these relays into the efficacy of our cell balancing strategy and its implications for
by sending appropriate activation signals. The IGBT relays are selected optimizing operational efficiency and prolonging the battery system's

8
S.L. Chavan, M.A. Kanawade and R.S. Ankushe Next Energy 8 (2025) 100258

Fig. 13. Characteristic with proportional-integral controller. SOC = state of charge.

Fig. 14. Characteristic with proportional-integral controller. SOC = state of charge.

longevity. The subsequent tables delineate the raw data before and after display of results before and after cell balancing is shown in Fig. 10 and
the cell balancing process, encapsulating the tangible outcomes of the Fig. 11, respectively.
hardware experiments. Table 4 shows circuit condition before cell The arrangement of these cells indicates that, with a given minimum
balancing where Fig. 8 and Fig. 9 shows circuit conditions before and SOC reference established by cell 3, cells 1 and 2, with higher SOC and
after cell balancing respectively. error percentages, are likely to discharge to meet the minimum SOC
In the context of the provided readings for 3 cells, each equipped level. The active engagement of their relays suggests ongoing discharge
with voltage measurements, SOC, error percentage, and relay operation processes, aiming to balance the entire battery pack. This strategic ar­
status, Acrucial insights into the battery system performance are re­ rangement ensures that cells with higher SOC and errors are actively
vealed. Cell 3 stands out with a minimum voltage of 3.03 V, indicating a contributing to the discharge process, aligning to achieve a balanced
75.73% minimum SOC with no observed error, and its switch is dis­ state while preventing over-discharge of the cells within the system.
engaged ("OFF"). This configuration suggests that Cell 3 is currently in a These readings collectively provide valuable information on the dy­
balanced state, and its switch remains OFF to prevent unnecessary namic interplay between cells and offer insights into the discharge
discharge. In contrast, Cell 1 exhibits a higher voltage of 3.36 V, cor­ strategy employed for maintaining optimal battery health. The Table 5
responding to a higher SOC of 83.94%, albeit with a 2.70% error, and shows circuit condition after cell balancing.
its switch is actively engaged ("ON"). Similarly, Cell 2 presents a high From Table 5, the updated readings for the 3 cells reveal a harmo­
voltage reading of 3.80 V, indicative of a high SOC of 94.68%, coupled nized state with synchronized voltage levels, SOC, and minimal error
with a 3.30% error, and its switch is also in an active state ("ON"). The percentages. Cell 3 now exhibits a voltage of 3.03 V, corresponding to a

9
S.L. Chavan, M.A. Kanawade and R.S. Ankushe Next Energy 8 (2025) 100258

Fig. 15. Characteristic without proportional-integral controller. SOC = state of charge.

Fig. 16. Characteristic without proportional-integral (PI) controller. SOC = state of charge.

75.73% SOC and, remarkably, a 0.00% error. The switch for Cell 3 is in "OFF" status of switches implies that the cells are in a stable equili­
the "OFF" state, indicating that this cell has reached a balanced condi­ brium, having discharged or charged as needed during the balancing
tion. Similarly, Cells 1 and 2 both display a voltage of 3.03 V, 75.73% process. This synchronized state aligns intending to achieve equalized
SOC, and 0.00% error, with their respective switches also in the "OFF" charge distribution among the cells, promoting optimal performance
position. This uniformity in voltage and SOC, coupled with the absence and longevity of the battery pack. The updated readings reflect the
of error (i.e. 0 error), signifies a successful cell balancing operation. The effectiveness of the cell balancing strategy employed, showcasing a

10
S.L. Chavan, M.A. Kanawade and R.S. Ankushe Next Energy 8 (2025) 100258

harmonized and balanced battery system ready for various operational From the above graph, it is conclude that cell having highest SOC is
demands. taking larger time to settle or balance. The above response is with PI
Corresponding to the results shown in Table 1, the lowest SOC is for controller where Kp and Ki values to settle the response are 0.65 and
cell number 2 which is 80, so for cell balancing, cell number 1 and cell 0.223, respectively
number 3 discharges through corresponding resistances to achieve SOC In Fig. 15, it is shown that a straight red line is reference SOC where
of nearly 80. The time required by cell number 1 to achieve final SOC is the yellow line and blue line.
19 sec where as time required by cell number 3 is 11 sec. Fig. 12 shows indicates that before cell balancing SOC was 90% and 100%, re­
variation of SOC with time for cell balancing. spectively. The cell having SOC of 90 is settled around 390 sec while
The equalization graph clearly shows the voltage levels of all cells cell having SOC of 100 is settled around 730 sec. From the above graph,
before, during, and after the balancing process. This helps in verifying it is conclude that cell having highest SOC is taking larger time to settle
that the balancing is effective and that all cells have reached the same or balance. The above response is without a PI controller.
voltage. The reunification results refer to the cells achieving uniform In Fig. 16, it is shown that straight yellow line is reference SOC
voltage, indicating successful balancing. where red line and blue line indicates that before cell balancing SOC
was 95% and 84%, respectively. The cell having SOC of 95 is settled
4.1. Loading effect on resistors and switching losses around 690 sec while cell having SOC of 84 is settled around 280 sec.
From the above graph, it is conclude that cell having highest SOC is
When a load is applied to the battery, the voltage drop across the taking larger time to settle or balance. The above response is without PI
resistors due to the load can be described by the equation: controller.

V_load = I_load × R. 5. Conclusion

The voltage drop can cause a shift in the balancing point, affecting
The presented research on simulation and hardware-based passive cell
how the resistors dissipate energy. In EVs, the load (current drawn by
balancing utilizing a PI controller has effectively tackled cell imbalance
the motor) varies with the speed of the vehicle. This variation impacts
within battery packs. Through the PI controller method, the presented
the balancing resistors and needs to be accounted for in the BMS design.
research meticulously determined the error between the lowest and
Switching losses occur when MOSFETs switch on and off are pro­
highest cell SOC, optimizing the operation of the Arduino-operated switch
portional to the voltage and current at the time of switching. The power
accordingly. With a minimum SOC set, the switch of the specified cell
loss during switching can be given by
remains OFF, while the switches for the other 2 cells remain ON until
P_switching = ½ × V_switch × I_switch × t_switch discharging through a resistor to the established minimum SOC.
The successful integration of the PI controller into the hardware
Where,
design has proven to be an efficient method for maintaining uniform
t_switch is the switching time. SOC levels across individual cells, thereby significantly extending the
overall lifespan and performance of the battery system. The presented
In passive cell balancing, the losses are mainly due to resistor dis­ research, marked by meticulous study, thoughtful design, and sys­
sipation. In active balancing (using inductors and capacitors), switching tematic experimentation, has demonstrated the feasibility and practi­
losses are more significant and must be minimized to improve effi­ cality of employing passive cell balancing techniques. The limitation of
ciency. the presented work is that even though results show the efficient way of
energy management between battery cells, the operating losses may
4.1.1. Load variation effect in the switching frequency affect the system efficiency. In order to reduce losses and get accurate
As the load varies (e.g., during acceleration or deceleration in a results proper tuning of the controller and abnormalities in the signals
vehicle), the switching frequency of MOSFETs in cell balancing circuits should be minimized with the use of an adaptive algorithm
may change. This affects the performance of the BMS and needs to be
optimized for different driving conditions. The BMS should adapt to the Declaration of Competing Interest
changes in load and maintain effective balancing without introducing
additional losses. When the SOC changes, the load on each cell may The authors declare that they have no known competing financial
change due to different discharge rates. This could lead to cell im­ interests or personal relationships that could have appeared to influ­
balance if not properly managed. The BMS should monitor SOC and ence the work reported in this paper.
adjust the balancing current accordingly to ensure uniform discharge
across all cells. References
The comparison between simulation and hardware results for 1 run
[1] S. Karmakar, T.K. Bera, A.K. Bohre, Novel PI controller and ANN controllers-based
are shown in Table 6. It clearly indicates that developed model works passive cell balancing for battery management system, IEEE Trans. Ind. Appl. 59 (6)
properly. The hardware results for this run are shown in Fig. 11. (2023) 7623–7634, [Link]
The balancing time under varying load conditions with and without [2] A.P. Shukla and R.A. Patel, Battery management system by passive cell balancing
for electric vehicle, international conference on power electronics &IoT applications
the PI controller is shown below from Tables 7,8 and 9.
in renewable energy and its control (PARC) (2022) Mathura, India, 1–6, doi: 10.
In Fig. 13, it is shown that straight red line is reference SOC where 1109/PARC52418.2022.9726646.
yellow line and blue line indicates that before cell balancing SOC was [3] H. Song, S. Lee, Study on the systematic design of a passive balancing algorithm
applying variable voltage deviation, Electronics 12 (12) (2023) 2587, [Link]
90% and 100%, respectively. The cell having SOC of 90 is settled
org/10.3390/electronics12122587.
around 300 sec while cell having SOC of 100 is settled around 600 sec. [4] W. Kurpiel, P. Deja, B. Polnik, M. Skóra, B. Miedziński, M. Habrych, G. Debita,
From the above graph it is conclude that cell having highest SOC is M. Zamłyńska, P. Falkowski-Gilski, Performance of passive and active balancing
taking larger time to settle or balance. The above response is with the PI systems of lithium batteries in onerous mine environment, Energies 14 (22) (2021)
7624, [Link]
controller. Where Kp and Ki values to settle the response is 0.8 and [5] V.-L. Pham, V.-T. Duong, W.A. Choi, Low cost and fast cell-to-cell balancing circuit
0.423, respectively. for lithium-ion battery strings, Electronics 9 (2) (2020) 248, [Link]
In Fig. 14, it is shown that straight yellow line is reference SOC 3390/electronics9020248.
[6] A. Khalid, A. Stevenson, A.I. Sarwat, performance analysis of commercial
where red line and blue line indicates that before cell balancing SOC passive balancing battery management system operation using a hardware-in-
was 95% and 84%, respectively. The cell having SOC of 95 is settled the-loop testbed, Energies 14 (23) (2021) 8037, [Link]
around 580 sec while cell having SOC of 84 is settled around 250 sec. en14238037.

11
S.L. Chavan, M.A. Kanawade and R.S. Ankushe Next Energy 8 (2025) 100258

[7] M. Kurucan, M. Özbaltan, Applications of artificial neural network based battery man­ [18] L. Maharjan, S. Inoue, H. Akagi, J. Asakura, Design and implementation of a PI-
agement systems: a literature review Renew, Sustain. Energy Rev. 192 (2024) 114262. controlled lithium ion battery charger with protection and monitoring, Eastern
[8] S. Aslam, Improvement of li-ion battery active balancer using PI-controller, J. Mediterranean University. (2020). Gazimağusa, North Cyprus.
Inform. Electr. Electron. Eng. 3 (2018) 1–10, [Link] [19] N. Samaddar, N.S. Kumar, R. Jayapragash, School of Electrical Engineering, Vellore
02.001. Institute of Technology, Chennai, India, Passive Cell Balancing of Li-ion batteries
[9] C.Y. Chun, B.H. Cho, and J. Kim, State-of-charge and remaining charge estimation used f or Automotive Applications, National Science, Engineering and Technology
of series-connected lithium-ion batteries for cell balancing scheme, IEEE Conference NCSET (2020).
International Telecommunications Energy Conference (INTELEC), Osaka, Japan, [20] Z.B. Omariba, L. Zhang, D. Sun, Review of battery cell balancing methodologies for
(2015) 1–5, doi:10.1109/INTLEC7572280. optimizing battery pack performance in electric vehicles, IEEE Access (2019),
[10] R. Paidi and S.K. Gudey, Active and passive cell balancing techniques for Li-ion bat­ [Link]
teries used in EVs, IEEE International Power and Renewable Energy Conference [21] R. Ling Member, Y. Dong, H. Yan, M. Wu, “Fuzzy-PI control battery equalization for
(IPRECON), Kollam, India, (2022) 1–6, doi: 10.1109/IPRECON55716.2022.10059573. series connected lithium-ion battery strings,” IEEE 7th International Power
[11] A. Panchal, K. Bhatt, S. Gitaye, M. Bhand and A. Sheikh, Design and Simulation of Electronics and Motion Control Conference – ECCE Asia (2012) Harbin, China.
an Inductor based Active Cell Balancing Circuit for Lithium-ion Batteries, 4th Global [22] S.W. Moore, P.J. Schneider, Delphi Automotive Systems, A review of cell equal­
Power, Energy and Communication Conference. (2022). ization methods for lithium ion and lithium polymer battery systems, Society of
[12] GPECOM, Nevsehir, Turkey, 2022, pp. 89–94, doi: 10.1109/GPECOM55404.2022. Automotive Engineers, Inc, 2001.
9815650. [23] S. El Fallah, J. Kharbach, J. Vanagas, Advanced state of charge estimation using
[13] A. Nath and B. Rajpathak, Analysis of cell balancing techniques in BMS for electric deep neural network, gated recurrent unit, and long short-term memory models for
vehicle, International Conference on Intelligent Controller and Computing for lithium-ion batteries under aging and temperature conditions, Appl. Sci. 14 (15)
Smart Power (ICICCSP), (2022) Hyderabad, India, pp. 1–6, doi: 10.1109/ (2024) 6648, [Link]
ICICCSP53532.2022.9862513. [24] J. Chen, Y. Zhang, W. Cheng, State of charge estimation for lithium-ion batteries
[14] R. Di Rienzo, M. Zeni, F. Baronti, R. Roncella, and R. Saletti, Passive balancing using gated recurrent unit recurrent neural network and adaptive Kalman filter, J.
algorithm for charge equalization of series connected battery cells, 2nd IEEE Energy Storage 55 (2022) 105396.
International Conference on Industrial Electronics for Sustainable Energy Systems [25] S. El Fallah, J. Kharbach, Z. Hammouch, State of charge estimation of an electric
(IESES), Cagliari, Italy, (2020), 73–79, doi: 10.1109/IESES45645.2020.9210643. vehicle’s battery using deep neural networks: simulation and experimental results,
[15] Y. Lee, S. Jeon, H. Lee, S. Bae, Comparison on cell balancing methods for energy J. Energy Storage 62 (2022) 6904.
storage applications, Indian J. Sci. Technol. 9 (17) (2016) 92316. [26] O. Demirci, S. Taskin, E. Schaltz, Review of battery state estimation methods for
[16] J. Cao, N. Schofield, A. Emadi, Battery balancing methods: a comprehensive review, electric vehicles - Part I: SOC estimation, J. Energy Storage (2024) 111435.
IEEE Vehicle Power and Propulsion Conference (VPPC), 2008, Harbin, China. [27] Y. Wang, C. Li, Q. Sun, Research on SOC estimation of lithium-ion batteries based
[17] M.M. Hoque, M.A. Hannan, A. Mohamed, Optimal algorithms for the charge on robust full order proportional integral observer, Int. J. Electrochem. Sci. 19 (8)
equalization controller of series connected lithium-ion battery cells in electric ve­ (2024) 100645.
hicle applications, IET Electrical Systems in Transportation.

12