Design and Analysis of Single-Inductor Power Converter For Both Battery Balancing and Voltage Regulation

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

2874 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 69, NO.

3, MARCH 2022

Design and Analysis of Single-Inductor Power


Converter for Both Battery Balancing and
Voltage Regulation
Gwangyol Noh , Student Member, IEEE, Jun Lee , Student Member, IEEE,
and Jung-Ik Ha , Fellow, IEEE

Abstract—In this article, a highly efficient battery- IS , IT Strong cell, target cell current.
balancing circuit with output voltage regulation for mo- VS , VT Strong cell, target cell voltage.
bile applications is proposed and analyzed. The proposed iBAL,pk , iBAL,avg Peak value, the average value of balanc-
circuit performs a cell-balancing operation for series-
connected multiple-battery cells and voltage regulation op- ing current.
eration simultaneously without additional power conver- tC1 , tL1 , tP 1 Time for high-side ON, low-side ON, and
sion stages. Therefore, it can be integrated into a bat- a sum of both in Mode 1.
tery pack with maximized power density and efficiency. tC2 , tL2 , tP 2 Time for high-side ON, low-side ON, and
These features are essential for recent low-power applica- a sum of both in Mode 2.
tions with multiple batteries, such as wireless/smart speak-
ers, drones, and electronic point-of-sale, that require high DM 1 , DM 2 Time duration of Modes 1 and 2.
power density. The operational principle, electrical model- SOCS , ΔSOCS Source cell SOC and its variance during
ing, and tradeoff relationship between the system power balancing operation.
losses and balancing characteristics are derived, and the SOCT , ΔSOCT Target cell SOC and its variance during
optimal design candidates of the proposed topology are balancing operation.
discussed in terms of Pareto optimization. The effective-
ness of the proposed system is verified with a proto- SOCi Intermediate cell SOC.
type board using GaN switches, and 96.05% efficiency is PL , PT , PS Load power, target cell power, and source
achieved with a 1.8 V, 0.5 A load. cell power.
Index Terms—Battery-balancing circuits, battery man-
M Conversion ratio between the source cell
agement systems (BMSs), output voltage regulation, power VS and the output voltage VO .
conversion circuits. RBAT , RSW Internal resistance of the single battery
cell, ON-resistance of the switch.
PBAT,1 , PCOND,1 Conduction loss due to battery internal
NOMENCLATURE
resistance, switch resistance in Mode 1.
SS1,T − SS3,T Top switch for each cell selection. POV,1 , PSW,1 Overlap loss, switching loss in Mode 1.
SS1,B − SS3,B Bidirectional bottom switch for each cell PBAT,2 , PCOND,2 Conduction loss due to battery internal
selection. resistance, switch resistance in Mode 2.
SP 1 − SP 3 High-side switch for each cell. POV,2 , PSW,2 Overlap loss, switching loss in Mode 2.
SL Low-side switch.
VP Pole voltage with respect to battery I. INTRODUCTION
ground.
ITH the rapid development of mobile, wearable devices,
VO
iL1 , iL
Output differential voltage.
Output inductor, load current. W and other commercial battery-powered applications, de-
mand for secondary batteries is on the rise. In particular, a
Manuscript received November 23, 2020; revised February 16, 2021; battery-balancing circuit is essential for systems with multiple
accepted March 2, 2021. Date of publication March 17, 2021; date of batteries connected in series since it contributes to maximizing
current version December 6, 2021. This work was supported in part by the utilization capacity of the battery energy. Typically, battery
Seoul National University Electric Power Research Institute, in part by
Brain Korea 21 Plus Project, and in part by the Institute of Engineering balancing is performed using either passive balancing or active
Research at Seoul National University. (Corresponding author: Jung-Ik balancing [1]–[4]. For passive balancing, excess energy is dis-
Ha.) sipated to the heat with a pair of shunt-connected switches and
The authors are with the Department of Electrical and Computer
Engineering, Seoul National University, Seoul 08826, South resistors. In order to reduce loss during a balancing operation,
Korea (e-mail: gwangyol.noh@gmail.com; leejun1672@snu.ac.kr; an active balancing method is proposed [2], [3], [5]–[7]. It
jungikha@snu.ac.kr). transfers energy between the cells using switches and passive
Color versions of one or more figures in this article are available at
https://doi.org/10.1109/TIE.2021.3065606. components, such as capacitors, inductors, and transformers.
Digital Object Identifier 10.1109/TIE.2021.3065606 Active balancing methods are classified into several categories
0278-0046 © 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.

Authorized licensed use limited to: UNIVERSITY OF NOTTINGHAM. Downloaded on February 23,2022 at 10:19:22 UTC from IEEE Xplore. Restrictions apply.
NOH et al.: DESIGN AND ANALYSIS OF SINGLE-INDUCTOR POWER CONVERTER 2875

Fig. 2. Proposed battery-balancing circuit with a voltage regulation for


three series-connected battery cells. In this study, three cells are used,
and the analysis and operation performed are all applicable to a system
with more cells.
Fig. 1. Typical power, energy, and signal flow of a battery pack con-
sisting of multicell batteries and a BMS [18]. The electrical requirements
of the power conversion circuit are increasing [19], which limits the
power density of the entire battery-powered system. In this work, the 2) The detailed analysis of the balancing operation is not
battery-balancing and voltage regulation circuits are combined into a
single circuit. considered [22].
3) The implementation of functions (battery balancing and
power conversion) is only covered without study or inter-
depending on the components. The switched capacitor method pretation on how the performance (efficiency and balanc-
[8] is advantageous for integration into a system because no ing speed) is achieved and optimized [19].
magnetic components are used. However, the equalization speed 4) Transformer-type topology (flyback) limits the power
is relatively slow compared with the methods using magnetic density, especially in the low-power, low-voltage
components [9]–[11]. Balancing methods using an inductor or applications [19].
transformer are widely used since the equalization speed is As described above, previous studies focused on the imple-
fast and the efficiency is high. However, the occupied area is mentation of the functions by relying on the conventional bal-
relatively large and the control is complex. Also, when using ancing performance. This article provides the analysis, tradeoff
multiwinding transformers, the design of leakage inductance relationships, and system optimization of the single-inductor
with high accuracy of the secondary winding is required, and the power converter for both battery balancing and voltage regula-
overall system becomes complicated and the size is increased tion through the consideration of the balancing time, balancing
[3], [10]. current, and power loss. To perform both battery balancing and
Recently, in many emerging applications, such as wire- voltage regulation, the circuit, as shown in Fig. 2, is considered
less/smart speakers, drones, and electronic point-of-sale (EPOS) for a low-power multiple-battery system. The topology, with a
[12]–[14], the use of multiple batteries is on the rise to shorten the single-inductor power converter, is a good candidate to achieve
charging time or due to the increase of loads (display, printing, high efficiency and fast balancing speed with a small area/size for
low-power motor drive, etc.) that require a high supply voltage low-power applications. This study is an extension of a previous
[12], [14]–[16]. conference paper [23] that introduces the optimal design of the
Fig. 1 shows a typical power, energy, and signal flow of a proposed circuit.
battery pack consisting of multicell batteries and its battery In the proposed topology (see Fig. 2), the bidirectional
management system (BMS) [17]. A BMS has many features, switches are used to select a cell to transfer energy back and
as shown in Fig. 1; thus, a power conversion circuit for var- forth. The load side consists of an inductor L1 and a capacitor C1
ious functions is required to properly operate a BMS. Since that supply power to the load and deliver energy to the target cell
the aforementioned low-power emerging applications require (weak cell). The detailed descriptions are covered in Section II.
small volumes and low weights, the power conversion circuits In Section III, the balancing characteristics with power conver-
limit the power density due to the magnetic components [15], sion are modeled, and the relationship between the balancing
[18]. Therefore, the optimal design of the battery-balancing time and state of charge (SOC) value versus the balancing
circuits and power conversion circuits is a key for achieving current is shown. The loss modeling of the proposed circuit is
a high-performance BMS [19]. discussed in Section IV, and the optimal design of the proposed
For this reason, as shown in Fig. 1, there have been studies circuit is performed in Section V with the loss–balancing time
that have achieved both battery-balancing operations and power Pareto optimization. In Section VI, three design candidates are
conversion (output voltage regulation) [19]–[22]. However, to verified with a prototype board and then compared, presenting
the best of our knowledge, previous studies have focused on the experimental results. In Section VII, discussions of previous
the functional implementation rather than studies that provide works are drawn. Finally, Section VIII concludes this article.
in-depth analysis and optimized design. Previous related studies
have the following limitations and disadvantages.
II. OPERATIONAL PRINCIPLE
1) Due to the lack of an analytical approach, similar perfor-
mance and efficiency are not guaranteed for the different In this section, the operational principle of the proposed
operating conditions [19]–[22]. topology is described. As shown in Fig. 3, the proposed circuit

Authorized licensed use limited to: UNIVERSITY OF NOTTINGHAM. Downloaded on February 23,2022 at 10:19:22 UTC from IEEE Xplore. Restrictions apply.
2876 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 69, NO. 3, MARCH 2022

Fig. 4. Output voltage and current ripple according to the duty ratio
according to input type for different battery voltages. With the same
output voltage, same inductor, and same operating frequency, the cell
input is advantageous for high efficiency due to the smaller current
ripple. For pack input, three cells are assumed.

Thus, in terms of the balancing speed and power conversion


efficiency, it can be understood that the cell input is more
advantageous for the efficient balancing and power conversion
than the pack input.
In this study, the battery-balancing and power conversion
operations are performed through a direct C2C operation, a
method specifically referred to as a cell-to-load-to-cell (C2L2C)
Fig. 3. Operation mode of the proposed topology. (a) Pack-powering operation. For C2L2C operation, the load-powering mode [see
mode. (b) Load-powering mode. (c) Cell-powering mode; in this opera-
tion mode, for example, B2 is selected.
Fig. 3(b)] and the cell-powering mode [see Fig. 3(c)] are applied.
Typical waveforms of the C2L2C operation for the battery
balancing and voltage regulation of the proposed topology are
has three operating modes depending on where the equalization shown in Fig. 5. For the following description, the direction from
current flows. In the pack-powering mode, as shown in Fig. 3(a), the battery to the load is considered to be the positive direction
the current flows from the pack to the load or vice-versa. In the of the current, as shown in Fig. 2. VO is well regulated with a
load-powering mode, the transferred energy is used in the load voltage ripple ΔV , which is modified from a typical burst-mode
or reserved in the load capacitor C1 for delivering back to the operation [24]. The battery voltage is constant during each mode,
cell or pack, as shown in Fig. 3(b). In the cell-powering mode, and the peak value of the inductor current iL1 is well regulated
as shown in Fig. 3(c), the current flows from the cell to the with iBAL,pk .
load or vice-versa. In each mode, the direction of the current is Fig. 5 shows that the C2L2C operation is divided into two
determined via appropriate control of switches. modes: Mode 1 and Mode 2, and the cell-to-load (C2L) mode
As shown in Fig. 3, the proposed topology provides three types and load-to-cell (L2C) mode, respectively. The circuit operates
of balancing operations: pack-to-cell (P2C), cell-to-cell (C2C), in Mode 1 until the output voltage VO reaches the high threshold;
and cell-to-pack (C2P). Among three possible operations, stud- then, it remains in Mode 2 until the low threshold is reached. In
ies on the balancing performance have shown that C2C type, the case of Mode 1, the power is transferred from the source cell
particularly direct C2C type, has faster equalization speeds than to the load with a positive inductor current iL1 , and in Mode 2,
the other types [2], [7]. the power is transferred from the load to the target cell with a
From the perspective of the power conversion, there are two negative inductor current iL1 .
types of conversion depending on the power input: pack or cell. The operating period of Mode 1 tP 1 is expressed as
Considering conduction loss I 2 R, a pack input has a larger iBAL,pk iBAL,pk
resistance than a cell input because the pack is composed of tP 1 = tC1 + tL1 = L1 + L1 . (1)
VS − VO VO
series-connected cells. Additionally, as shown in Fig. 4, for all
ranges of battery voltages, the current ripple of the cell input is Likewise, the operating period of Mode 2 tP 2 is expressed as
less than that of the pack input. A smaller current ripple yields follows:
a smaller rms current, which means that the power loss for the iBAL,pk iBAL,pk
tP 2 = tL2 + tC2 = L1 + L1 . (2)
cell type is smaller. VO VT − VO

Authorized licensed use limited to: UNIVERSITY OF NOTTINGHAM. Downloaded on February 23,2022 at 10:19:22 UTC from IEEE Xplore. Restrictions apply.
NOH et al.: DESIGN AND ANALYSIS OF SINGLE-INDUCTOR POWER CONVERTER 2877

Fig. 6. Each SOC profile of the proposed topology (C2L2C operation)


versus time is divided into two cases according to the conditions in (7).
(a) Case I: the SOC of an intermediate cell is between the source cell
SOC and the final SOC. (b) Case II: the SOC of an intermediate cell is
between the final SOC and the target cell SOC.
Fig. 5. Operational waveforms of the battery balancing with voltage
regulation for a C2L2C operation. In this waveform, B3 is a source
cell, and B1 is a target cell. Mode 1 operates by repeating the pair of
the cell-powering mode (see Fig. 3(c), B3 in this waveform) and the The relationship between the battery voltage and SOC is as-
load-powering mode. Mode 2 operates by repeating the pair of the sumed to be a simple linear model for providing insights into
load-powering mode and cell-powering mode (see Fig. 3(c), B1 in this
waveform). tC1 , tL1 , tC2 , and tL2 follow (1) and (2). the balancing characteristics of the proposed topology.
Fig. 6 shows the SOC plot for a typical C2L2C operation of
the proposed topology. First, the source cell delivers energy to
Notably, the peak value of the inductor current iBAL,pk is the load side. Then, the load side consists of the inductor L1 ,
an important design element of the proposed circuit, which is and the capacitor C1 delivers the energy to the target cell. In
discussed in detail in Section III. this operation, cells with the highest SOC and the lowest SOC
The change of the output voltage in each mode is assumed to are selected as the source cell and the target cell, respectively.
be linear, and the voltage changes in both modes are the same Then, the charge/discharge variance for time t of each cell is as
in the steady state; therefore, the time duration of each mode is follows:
expressed as follows:  t
t1 − t0 0.5iBAL,pk + iL ΔSOCS (t) · QBAT = IS (t) dt = iBAL,avg · t (5)
DM 1 ≡ = (3) 0
t2 − t0 iBAL,pk  t
t2 − t1 0.5iBAL,pk − iL ΔSOCT (t) · QBAT = α {μB iBAL,avg − M iL (t)} dt.
DM 2 ≡ = . (4) 0
t2 − t0 iBAL,pk (6)

III. BALANCING CHARACTERISTICS Here, μB is the efficiency of the proposed circuit, which
In this section, the balancing characteristics of the proposed is μB = (PL + PT )/PS , where PL , PT , and PS are the load
topology according to the balancing current and load current power, target cell power, and source cell power, respectively.
are described. The balancing characteristics, including the final μB in (6), according to the battery voltage, is calculated by
SOC value and the balancing time of the proposed circuit, are (14)–(23). M is the conversion ratio between the source cell VS
considered; then, the power loss model is covered in Section IV. and the output voltage VO , M = VO /VS . α is the ratio between

Authorized licensed use limited to: UNIVERSITY OF NOTTINGHAM. Downloaded on February 23,2022 at 10:19:22 UTC from IEEE Xplore. Restrictions apply.
2878 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 69, NO. 3, MARCH 2022

the source cell VS and the target cell VT , α VT = VS . QBAT is


the capacity of the battery.
Considering three or more cells, the balancing of the highest
or lowest SOC converges to the same SOC as the intermediate
cell SOCi (0). In other words, tx is obtained from the following
equations:

⎨ SOCi (0) = SOCS (0) − ΔSOCS (tx,a )
SOCi (0) = SOCT (0) + ΔSOCT (tx,b ) . (7)

tx = min (tx,a , tx,b )
In Fig. 6(a), since the minimum value of tx is determined by
the source cell, tx is equal to tx1 . Likewise, in Fig. 6(b), tx is
determined by the target cell, so tx is equal to tx2 . In other words,
Fig. 7. Plot of the final SOC and balancing time as a function of the
the total balancing time tBAL is divided into before and after the load current and balancing current; the final SOC increases with an
conditions in (7) are satisfied, i.e., tx1 and ty1 or tx2 and ty2 , increase in the balancing current and tends to decrease with an increase
as shown in Fig. 6. Before satisfying the conditions in (7), each in the load current. The balancing time tends to decrease with increasing
balancing current and increasing load current. The initial SOC values are
cell is balanced with the rate of change of (5) and (6). After the the same as the values of the experiment (see Fig. 17).
condition is satisfied, the balancing rate is changed according to
the relationship between each cell.
For Case I of Fig. 6(a), when iBAL (t) and iL (t) are constant IV. LOSS MODELING
within the time period, the total balancing time is expressed as This section describes the loss modeling of the proposed
follows: circuit. The losses of each mode are divided into the following:
1) battery conduction loss due to the internal resistance of the
tBAL1 = tx1 + ty1 (8)
battery cell;
(SOCS (0) − SOCi (0)) 2) switch conduction loss due to the ON-resistance of the
tx1 = · QBAT (9)
iBAL,avg power switch;
3) switch overlap loss;
SOCS (tx1 ) − SOCT (tx1 )
ty1 = · QBAT 4) switching loss due to the output capacitance of each switch.
0.5iBAL,avg + α {μB iBAL,avg − M iL } Modeling for each loss factor is derived based on the opera-
(10) tional principle, as described in Section II, and the time average
where tx1 and ty1 are the time durations of each, as shown value of each loss component is expressed to determine the total
in Fig. 6(a). Likewise, the total balancing time for Case II is loss.
expressed as Before proceeding with the loss analysis, the detailed switch
operation of C2L2C is described. For both Modes 1 and 2, the
tBAL2 = tx2 + ty2 (11) operation is performed with four high-side switches and one
low-side switch SL , as shown in Fig. 2. For example, three of the
(SOCi (0) − SOCT (0))
tx2 = · QBAT (12) four high-side switches SS1T and SS1B (SS1B represents two
α {μB iBAL,avg − M iL }
switches) are responsible for the cell selection, and the other SP 1
SOCS (tx2 ) − SOCT (tx2 ) is responsible for the pulsewidth modulation (PWM) switching
ty2 = · QBAT .
iBAL,avg + 0.5α {μB iBAL,avg − M iL } paired with the low-side switch SL for B1 .
(13) Thus, the battery conduction loss for Mode 1 is expressed as
  2
As shown in Fig. 6, the SOC value at tBAL is determined 2 tC1
PBAT,1 = iBAL,rms RBAT = iBAL,pk RBAT
by both the balancing current and the load current. Considering 3tP 1
the two extreme cases, if the balancing current is very large (14)
compared with the load current, thefinal SOC (SOCFIN ) is the where RBAT is the internal resistance of the single battery cell.
average value of each initial SOC ( SOC(0)/3), which is the The switch conduction loss for Mode 1 is expressed as
achievable maximum value of the final SOC. On the other hand, PCOND,1
if the balancing current is very small compared with the load
  2   2
current, then the final SOC converges to SOCT (0), which is the tC1 tL1
minimum of the final SOC values. = 4 iBAL,pk + iBAL,pk RSW
3tP 1 3tP 1
Fig. 7 shows a plot of the final SOC value and balancing
(15)
time with respect to the balancing and load currents for the
operating conditions to be used in this study. The final SOC where RSW is the ON-resistance of the switch. Since the battery
value tends to increase as the balancing current increases and cell transition period is longer than the PWM switching period,
tends to decrease as the load current increases. However, the i.e., tP 1  (t2 − t0 ), the PWM pairs (SP 1 ∼ SP 3 , and SL ) form
balancing time becomes shorter as both currents increase. dominant switching losses. Thus, the switch overlap loss and the

Authorized licensed use limited to: UNIVERSITY OF NOTTINGHAM. Downloaded on February 23,2022 at 10:19:22 UTC from IEEE Xplore. Restrictions apply.
NOH et al.: DESIGN AND ANALYSIS OF SINGLE-INDUCTOR POWER CONVERTER 2879

Fig. 8. Plot of the total power loss versus the balancing current with Fig. 9. Plot of the total power loss versus various inductances.
4.7 µH of inductance; the switching losses are dominant as the balanc-
ing current decreases, and the conduction losses are dominant as the
balancing current increases.
is observed that the switching losses become dominant as the
balancing current decreases, and the conduction losses become
switching loss for Mode 1 are expressed as dominant as the balancing current increases. Thus, the balancing
current iBAL,pk directly determines the total power loss, which
0.5iBAL,pk VS
POV,1 = 2 Ttr (16) is an important factor in the overall design process.
tP 1
0.5COSS VS2 V. PARAMETER DESIGN AND SYSTEM OPTIMIZATION
PSW,1 = 2 (17)
tP 1 In this section, the effect of each design consideration on
where VS is the voltage of the source cell, Ttr is the ON/OFF tran- the overall performance is investigated, and the optimal design
sition time of each switch, and COSS is the output capacitance candidates are extracted via system optimization.
provided from the datasheet of the component.
Therefore, the total power loss for Mode 1 is expressed as A. Output Inductor and Capacitor Design
follows:
As shown in (1) and (2), the operating frequency of the
P1 = (PBAT,1 + PCOND,1 + POV,1 + PSW,1 ) DM 1 . (18) proposed circuit is directly affected by the output inductance
L1 . Specifically, the change in the inductance is closely related
In a similar manner, the power loss for Mode 2 is expressed to the switching loss. Since the proposed circuit is integrated into
as follows: the battery pack of the mobile and consumer devices, such as
  2
tL2 wireless/smart speakers, drones, and EPOS, the output inductor
PBAT,2 = i2BAL,rms RBAT = iBAL,pk RBAT must be small in size with small low-frequency copper loss. In
3tP 2
this study, a set of commercial inductors with the same package
(19)
size is selected [25]. For a given package size, an increase in

  2
tL2 inductance would result in a larger low-frequency copper loss
PCOND,2 = 4 iBAL,pk since only the number of coil turns increases. As shown in Fig. 9,
3tP 2
an increase in inductance L1 reduces the switching loss and
  2 increases the conduction loss, as mentioned above. In this study,
tC2
+ iBAL,pk RSW (20) 4.7 μH of inductance was selected considering both the size and
3tP 2
power loss of the output inductor for the desired operation of the
0.5iBAL,pk VT proposed circuit.
POV,2 = 2 Ttr (21) For the output capacitor C1 , it is required to transfer the energy
tP 2
to the target cell from the load side. Thus, the output capacitor
0.5COSS VT2 C1 is designed to satisfy the following condition:
PSW,2 = 2 (22)
tP 2
C1 > L1 i2BAL,pk /ΔV 2 . (24)
P2 = (PBAT,2 + PCOND,2 + POV,2 + PSW,2 ) DM 2 .
(23) Here, the ripple of the voltage ΔV and the capacitor have a
tradeoff relationship. As the size of the capacitor C1 increases,
The total loss P1 + P2 versus the balancing current is illus-
the ripple of the voltage is reduced. Thus, ΔV and C1 are
trated in Fig. 8. The conduction loss represents the sum of PBAT
required by the given specification of a system.
and PCOND , and the switching loss represents the sum of POV
and PSW . As in (1) and (2), the operating frequency and duty
B. Effect of Balancing Current
ratio are closely related to iBAL,pk . In other words, the trend of
power loss, as shown in Fig. 8, includes the tendency of not only Based on the modeling analyzed in Section IV, the current-
conduction loss but also switching loss according to iBAL,pk . It dependent loss and the balancing time are shown in Fig. 10.

Authorized licensed use limited to: UNIVERSITY OF NOTTINGHAM. Downloaded on February 23,2022 at 10:19:22 UTC from IEEE Xplore. Restrictions apply.
2880 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 69, NO. 3, MARCH 2022

Fig. 10. Plot of the tradeoff region between the power loss and bal- Fig. 11. Pareto front of the proposed circuit and three selected design
ancing time for the various balancing currents and load currents (4.7 µH candidates.
of inductance). The initial SOC values are the same as the values of the
experiment (see Fig. 17).

TABLE I
PARAMETER VALUES SUMMARY

Fig. 12. Photograph of the prototype board of the proposed circuit with
a three series-connected battery string.

parameter is normalized to its maximum value in the operating


Despite the change in the load current, the power loss remains area with the design constraints.
at a similar level because the duty ratios in Mode 1 and Mode 2 In Fig. 11, the Pareto front illustrates the tradeoff relationship
(DM 1 and DM 2 ) have the opposite direction to the load current observed in Fig. 10. Considering the safe region of the battery
IL , which is observed in (3) and (4). Therefore, the total loss used in this work, three design candidates are selected to validate
does not change significantly by the load current. this optimization process, and the balancing current iBAL,pk for
The relationship between the loss and balancing current is designs I, II, and III is 1.25, 1.05, and 0.65 A, respectively.
illustrated, as shown in Fig. 8, in Section IV. As observed in
Fig. 7, the balancing time decreases as the balancing current VI. EXPERIMENTAL RESULTS
(or load current) increases. In other words, when operating in
the gray zone of Fig. 10 with the designed values, a tradeoff The prototype board for verifying the proposed circuit and the
relationship between the power loss and balancing performance analysis performed in this study is shown in Fig. 12. In this study,
is inevitable. The parameter values used in this analysis are given GaN switches and gate drivers are used to minimize the power
in Table I. loss and component area. The board consists of a bidirectional
switch set, output inductor L1 , output capacitor C1, and sensing
circuit. The component values used in the prototype board are
C. Design Optimization With the Pareto Front summarized in Table I.
The design parameters are selected to optimize both the Fig. 13(a) shows the flowchart of the proposed operation. At
balancing speed and system efficiency. Design factors, such as the start of the operation, the PWM duty ratio is set to zero
the balancing current, load current, output inductance, output and slowly increases until the target voltage is reached with the
capacitance, operating frequency, duty ratio, and output voltage source cell. When the output voltage reaches the target value,
ripple, are taken into account for the optimal design of the the controller determines whether the balancing operation is re-
proposed circuit. Based on the analysis performed in Section quired by sensing the battery voltages. If the balancing operation
V-B, the balancing time and the total power loss are optimized is required, the operating frequency and duty ratio for each mode
by exploring the Pareto front [26], [27]. The Pareto front is the set are calculated by (1) and (2) to operate with targeted iBAL,pk .
of feasible optimum outputs. The equations done in Sections III If the tolerance is smaller than a certain value, then only the
and IV are coded and explored by using the fgoalattain function voltage regulation is performed without a balancing operation
in the MATLAB optimization toolbox [28]. In this work, each via C2L power conversion. Since the balancing current iBAL,pk

Authorized licensed use limited to: UNIVERSITY OF NOTTINGHAM. Downloaded on February 23,2022 at 10:19:22 UTC from IEEE Xplore. Restrictions apply.
NOH et al.: DESIGN AND ANALYSIS OF SINGLE-INDUCTOR POWER CONVERTER 2881

Fig. 15. Waveforms of the output voltage VO , pole voltage VP , and


inductor current IL1 for the C2L2C mode of the proposed circuit; two high
SOC cells (B1 and B2 ) have similar values as the balancing operation
proceeds. B1 and B2 alternately perform as the source cell and transfer
the energy to B3 . The output load voltage VO is stable despite cell
changes.

output voltage regulation, the burst mode with a transition be-


tween Modes 1 and 2 is implemented. The operation frequency
and its duty ratio follow the relations in (1)–(4). In Mode 1,
B1 is selected as the source cell, and the energy is transferred
to the load. In Mode 2, the energy stored in the load capacitor
C1 is delivered through inductor L1 to target cell B3 . The mode
transition can be observed through the change in the pole voltage
VP and the direction of the inductor current iL1 , as shown in
Fig. 5.
Fig. 15 shows that two high energy cells (B1 and B2 ) have
similar values as the balancing operation proceeds. B1 and B2
alternately perform as source cells and transfer the energy to B3 .
The output voltage VO is well regulated despite cell changes.
Fig. 13. (a) Flowchart of the proposed operation. (b) System configu- Additionally, the operation can be observed through the pole
ration of the proposed circuit. voltage VP and the inductor current iL1 .
The high-frequency oscillations at the end period are shown
in Figs. 14 and 15. In this study, the corresponding power loss
is minimized through the proper selection of the GaN switches
and the minimum parasitic inductance PCB design [29].
Fig. 16 shows the measured SOCs over time under an unequal
initial SOC distribution. The experimental results are compared
with the MATLAB estimated results based on (5)–(13). The
overall balancing performance indicates the adequate accuracy
between the estimated and measured results.
Fig. 17 shows the measured efficiency and balancing time for
each design candidate. As explained in Section V, design I has the
shortest balancing time, and design III has the highest efficiency.
Fig. 14. C2L2C operation of the proposed circuit. As observed, the
output voltage operates in the burst mode, as described in Section V,
The validity of the analysis and optimization process performed
and repeats the C2L mode (B1 to load) and L2C mode (load to B3 ) in the study is confirmed based on the quantitative differences
within the voltage window. The initial SOC of each cell is as follows: in the experimental results. As discussed in [24], the burst-mode
SOC1 = 90%, SOC2 = 65%, and SOC3 = 40%.
operation has nearly constant efficiency over the entire range of
the load current, and the burst mode for the proposed circuit is
also validated by the experimental results.
is calculated by (1) and (2), the battery and output voltage are
required to be sensed, as shown in Fig. 13(b). The output voltage
is sensed and detected with an analog comparator, as shown VII. DISCUSSION
in Fig. 13(b). Each signal for the cell selection and PWM is The proposed topology is compared with the previous studies
properly generated according to the battery voltages by a digital of the articles presented in [19], [22], [30], and [31] in Table II,
controller (TMS320C28346, TI). which shows the balancing topology, converter type, size, effi-
The C2L2C operation is shown in Fig. 14. The output voltage ciency, and balancing speed. In [30], 2N − 2 of switches and
VO stably reaches the target voltage with the source cell. For the 2N − 1 of diodes are used. Thus, the decrease in efficiency is

Authorized licensed use limited to: UNIVERSITY OF NOTTINGHAM. Downloaded on February 23,2022 at 10:19:22 UTC from IEEE Xplore. Restrictions apply.
2882 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 69, NO. 3, MARCH 2022

TABLE II
COMPARISON WITH PREVIOUS WORKS

In the case of the switch, a unidirectional switch is counted as one switch.


C2P – Cell-to-pack, P2C – Pack-to-cell, P2L – Pack-to-load, C2L – Cell-to-load, C2C – Cell-to-cell, and C2L2C – Cell-to-load-cell.
The efficiency with a star(∗) refers to the balancing operation only.

Fig. 17. Measured efficiency and balancing time versus various load
currents, IL = 0.05, 0.2, and 0.5 A. The initial distribution of SOCs is as
follows: SOC1 = 50%, SOC1 = 65%, and SOC1 = 100%. According
to these experimental results, the selection of candidates based on the
analysis is well validated.

[19], a high degree of freedom in the balancing operation with


fast balancing speed is achieved by applying a bidirectional
flyback converter on each cell stage. However, each stage needs
a corresponding transformer, which dominates the size and
weight of the system. Moreover, unlike other studies, each stage
also requires the output current sensing with A/D conversion
for the current control method. This further increases the size
and complexity of the controller. In [22], 2N + 1 of switches
with multiwinding transformer are used for a medium size. The
balancing speed is limited due to the pack-to-load (P2L) type
balancing method, which relies only on discharging the load
current at each battery cell.
Compared with other studies [19], [22], [30], and [31], this
Fig. 16. Measured SOC profile during balancing operation of three
study provides many advantages. Due to the characteristics of
cells. The balancing time follows (8)–(13) depending on each condition the proposed topology, an additional power conversion circuit is
(the balancing current, load current, and initial SOCs). The initial SOC not required. A single-inductor power converter topology with
distribution of (a) SOC3 > SOC2 > SOC1 , (b) SOC3 > SOC1 > SOC2 ,
and (c) SOC1 > SOC2 >SOC3 .
optimization based on in-depth analysis achieves a smaller size
and higher efficiency. The increase in the number of switches due
to the bidirectional operation may increase the cost; however, the
severe due to large diode conduction loss. In [31], 4N + 2 of cost of the switch has been continuously decreasing recently.
switches and 2 of diodes are used with two transformers for Consequently, the proposed circuit performs both battery bal-
the step-up/down operation, which increases the size, area, ancing and power conversion (voltage regulation) with better
and weight of the circuit. Additionally, the balancing speed is performance in terms of the efficiency and balancing character-
limited because of the pack-type balancing method [2], [7]. In istics compared with those in previous studies.

Authorized licensed use limited to: UNIVERSITY OF NOTTINGHAM. Downloaded on February 23,2022 at 10:19:22 UTC from IEEE Xplore. Restrictions apply.
NOH et al.: DESIGN AND ANALYSIS OF SINGLE-INDUCTOR POWER CONVERTER 2883

VIII. CONCLUSION [15] “TPS62172 data sheet, product information and support.” [Online]. Avail-
able: https://www.ti.com/product/TPS62172, Accessed on: Nov. 2020.
In this article, we presented a high-efficiency balancing cir- [16] “TI designs non-military drone, robot, or RC 2S1P battery management
cuit with voltage regulation for low-power mobile applications, solution reference design.” [Online]. Available: https://www.ti.com/lit/
ug/tidubq4a/tidubq4a.pdf?ts=1611033709585&ref_url=https%253A%
which was advantageous for integration into battery packs. To 252F%252Fwww.google.com%252F, Accessed on: Nov. 2020.
optimize the proposed topology, the principle of operation, [17] H. Rahimi-Eichi, U. Ojha, F. Baronti, and M.-Y. Chow, “Battery manage-
balancing characteristics, and loss models were derived. The ment system: An overview of its application in the smart grid and electric
vehicles,” IEEE Ind. Electron. Mag., vol. 7, no. 2, pp. 4–16, Jun. 2013.
tradeoff relationship between the balancing performance and ef- [18] “LT1616 datasheet and product info | analog devices.” [Online]. Avail-
ficiency was analyzed and discussed. Then, three optimal design able: https://www.analog.com/en/products/lt1616.html, Accessed on:
candidates were selected through the Pareto front. This study Nov. 2020.
[19] Y.-D. Yang, K.-Y. Hu, and C.-H. Tsai, “Digital battery management
focused on the quantitative analysis of a C2L2C operation with design for point-of-load applications with cell balancing,” IEEE Trans.
a burst mode, which was distinguished from previous studies. Ind. Electron., vol. 67, no. 8, pp. 6365–6375, Aug. 2020.
To verify the analysis performed in this work, a GaN-based [20] Y. C. Hsieh, S. P. Chou, and C. S. Moo, “Balance discharging for series-
connected batteries,” in Proc. IEEE 35th Annu. Power Electron. Spec.
prototype board was implemented. The results showed that the Conf., Jun. 2004, vol. 4, pp. 2697–2702.
proposed circuit operated with high efficiency up to 96.05% [21] M. Evzelman, M. M. U. Rehman, K. Hathaway, R. Zane, D. Costinett,
and a high-speed balancing operation of 2270 s (37.8 min) for and D. Maksimovic, “Active balancing system for electric vehicles with
incorporated low-voltage bus,” IEEE Trans. Power Electron., vol. 31,
three series-connected battery cells (2000 mAh each cell). The no. 11, pp. 7887–7895, Nov. 2016.
outstanding performance and high functionality can minimize [22] M. Shousha, T. McRae, A. Prodić, V. Marten, and J. Milios, “Design and
system redundancy and achieve a high power density for battery- implementation of high power density assisting step-up converter with
integrated battery balancing feature,” IEEE J. Emerg. Sel. Topics Power
powered low-power mobile devices. Electron., vol. 5, no. 3, pp. 1068–1077, Sep. 2017.
[23] G. Noh and J.-I. Ha, “High reliable power conversion system with active
REFERENCES battery balancing capability,” in Proc. IEEE Appl. Power Electron. Conf.
Expo., Mar. 2020, pp. 3358–3363.
[1] B. Lindemark, “Individual cell voltage equalizers (ICE) for reliable battery [24] J. Hu, A. D. Sagneri, J. M. Rivas, Y. Han, S. M. Davis, and D. J. Perreault,
performance,” in Proc. 13th Int. Telecommun. Energy Conf., Nov. 1991, “High-frequency resonant SEPIC converter with wide input and output
vol. 91, pp. 196–201. voltage ranges,” IEEE Trans. Power Electron., vol. 27, no. 1, pp. 189–200,
[2] S.-H. Park, K.-B. Park, H.-S. Kim, G.-W. Moon, and M.-J. Youn, “Single- Jan. 2012.
magnetic cell-to-cell charge equalization converter with reduced number [25] “Bourns—power inductors—AEC-Q200 compliant.” [Online]. Available:
of transformer windings,” IEEE Trans. Power Electron., vol. 27, no. 6, https://www.bourns.com/products/magnetic-products/power-inductors-
pp. 2900–2911, Jun. 2012. aec-q200-compliant/product/SRN5040TA, Accessed on: Nov. 2020.
[3] Y. Shang, B. Xia, C. Zhang, N. Cui, J. Yang, and C. C. Mi, “An auto- [26] T. M. Andersen et al., “Modeling and pareto optimization of on-chip
matic equalizer based on forward–flyback converter for series-connected switched capacitor converters,” IEEE Trans. Power Electron., vol. 32,
battery strings,” IEEE Trans. Ind. Electron., vol. 64, no. 7, pp. 5380–5391, no. 1, pp. 363–377, Jan. 2017.
Jul. 2017. [27] N. Rashidi, Q. Wang, R. Burgos, C. Roy, and D. Boroyevich, “Multi-
[4] Z. B. Omariba, L. Zhang, and D. Sun, “Review of battery cell balancing objective design and optimization of power electronics converters with
methodologies for optimizing battery pack performance in electric vehi- uncertainty quantification—Part I: Parametric uncertainty,” IEEE Trans.
cles,” IEEE Access, vol. 7, pp. 129335–129352, 2019, doi: 10.1109/AC- Power Electron., vol. 36, no. 2, pp. 1463–1474, Feb. 2021.
CESS.2019.2940090. [28] “Multiobjective optimization algorithms—MATLAB & simulink.”
[5] S. Yarlagadda, T. T. Hartley, and I. Husain, “A battery management [Online]. Available: https://www.mathworks.com/help/optim/ug/
system using an active charge equalization technique based on a dc/dc multiobjective-optimization-algorithms.html, Accessed on: Nov. 2020.
converter topology,” IEEE Trans. Ind. Appl., vol. 49, no. 6, pp. 2720–2729, [29] T. Liu, T. T. Y. Wong, and Z. J. Shen, “A survey on switching oscillations
Nov./Dec. 2013. in power converters,” IEEE J. Emerg. Sel. Topics Power Electron., vol. 8,
[6] T. H. Phung, A. Collet, and J.-C. Crebier, “An optimized topology for no. 1, pp. 893–908, Mar. 2020.
next-to-next balancing of series-connected lithium-ion cells,” IEEE Trans. [30] A. M. Imtiaz and F. H. Khan, “‘Time shared flyback converter’ based
Power Electron., vol. 29, no. 9, pp. 4603–4613, Sep. 2014. regenerative cell balancing technique for series connected Li-ion battery
[7] Y. Li, J. Xu, X. Mei, and J. Wang, “A unitized multiwinding transformer- strings,” IEEE Trans. Power Electron., vol. 28, no. 12, pp. 5960–5975,
based equalization method for series-connected battery strings,” IEEE Dec. 2013.
Trans. Power Electron., vol. 34, no. 12, pp. 11981–11989, Dec. 2019. [31] M. A. Hannan, M. M. Hoque, S. E. Peng, and M. N. Uddin, “Lithium-ion
[8] Y. Ye, K. W. E. Cheng, Y. C. Fong, X. Xue, and J. Lin, “Topology, battery charge equalization algorithm for electric vehicle applications,”
modeling, and design of switched-capacitor-based cell balancing systems IEEE Trans. Ind. Appl., vol. 53, no. 3, pp. 2541–2549, May/Jun. 2017.
and their balancing exploration,” IEEE Trans. Power Electron., vol. 32,
no. 6, pp. 4444–4454, Jun. 2017.
[9] M. Caspar, T. Eiler, and S. Hohmann, “Systematic comparison of active
balancing: A model-based quantitative analysis,” IEEE Trans. Veh. Tech-
nol., vol. 67, no. 2, pp. 920–934, Feb. 2018.
[10] Y. Shang, N. Cui, and C. Zhang, “An optimized any-cell-to-any-cell equal-
izer based on coupled half-bridge converters for series-connected battery Gwangyol Noh (Student Member, IEEE) re-
strings,” IEEE Trans. Power Electron., vol. 34, no. 9, pp. 8831–8841, ceived the B.S. and M.S. degrees in electri-
Sep. 2019. cal engineering from Sogang University, Seoul,
[11] K.-M. Lee, S.-W. Lee, Y.-G. Choi, and B. Kang, “Active balancing of South Korea, in 2009 and 2011, respectively. He
Li-ion battery cells using transformer as energy carrier,” IEEE Trans. Ind. is currently working toward the Ph.D. degree in
Electron., vol. 64, no. 2, pp. 1251–1257, Feb. 2017. electrical engineering with Seoul National Uni-
[12] “BQ25886 data sheet, product information and support.” [Online]. Avail- versity, Seoul.
able: https://www.ti.com/product/BQ25886, Accessed on: Nov. 2020. Since 2011, he has been a Staff Engineer
[13] “LT1505 datasheet and product info | analog devices.” [Online]. Avail- with Samsung Electronics Co., Ltd., Seoul,
able: https://www.analog.com/en/products/lt1505.html, Accessed on: where he is working on power management
Nov. 2020. IC and battery management IC. His current re-
[14] “BQ25790 data sheet, product information and support.” [Online]. Avail- search interests include high-efficient power electronics, high conver-
able: https://www.ti.com/product/BQ25790, Accessed on: Nov. 2020. sion ratio converters, and battery management systems.

Authorized licensed use limited to: UNIVERSITY OF NOTTINGHAM. Downloaded on February 23,2022 at 10:19:22 UTC from IEEE Xplore. Restrictions apply.
2884 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 69, NO. 3, MARCH 2022

Jun Lee (Student Member, IEEE) was born in Jung-Ik Ha (Fellow, IEEE) received the B.S.,
Seoul, South Korea, in 1993. He received the M.S., and Ph.D. degrees in electrical engineer-
B.S. degree in electrical engineering in 2015 ing from Seoul National University, Seoul, South
from Seoul National University, Seoul, South Ko- Korea, in 1995, 1997, and 2001, respectively.
rea, where he is currently working toward the From 2001 to 2002, he was a Researcher
Ph.D. degree in electrical engineering. with Yaskawa Electric Co., Kitakyushu, Japan;
His current research interests include elec- from 2003 to 2008, he was a Senior and Prin-
tric energy conversion, electric machine control, cipal Engineer with Samsung Electronics Co.,
and magnetic manipulation systems. South Korea; and from 2009 to 2010, he was
a Chief Technology Officer with LS Mecapion
Co., South Korea. Since 2010, he has been a
Professor with the Department of Electrical and Computer Engineering,
Seoul National University, Seoul. He is also working with Seoul National
University Electric Power Research Institute. From 2016 to 2017, he
was a Visiting Scholar with the Massachusetts Institute of Technology,
Cambridge, MA, USA. He has authored more than 150 papers pub-
lished in international journals and conference proceedings in the area
of power electronics and motor drives. His current research interests
include circuits and control in high efficiency and integrated electric
energy conversions for various industrial fields.
Dr. Ha is an Associate Editor for the IEEE Journal of Emerging and
Selected Topics in Power Electronics, and the Editor-in-Chief for the
Journal of Power Electronics.

Authorized licensed use limited to: UNIVERSITY OF NOTTINGHAM. Downloaded on February 23,2022 at 10:19:22 UTC from IEEE Xplore. Restrictions apply.

You might also like