Você está na página 1de 10

IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 3, NO.

4, DECEMBER 2015

1001

DC Active Power Filter-Based Hybrid Energy


Source for Pulsed Power Loads
Vladimir Yuhimenko, Student Member, IEEE, Chaim Lerman, Student Member, IEEE,
and Alon Kuperman, Senior Member, IEEE

Abstract In this paper, design of a capacitor semiactive


hybrid source for powering pulsed power loads based on
dc power filter principle is presented. The system consists
of an energy source connected directly to a load, supported
by a bidirectional buckboost dcdc converter interfaced
supercapacitor (SC). The converter is controlled such that the
SC supplies the dynamic component of the load power, leaving the
energy source to supply a near-constant power to satisfy average
load demand. The control algorithm is adopted from the power
filter theory, allowing to reduce the stress of an energy rich source
despite operating under a high-power demanding load. Moreover,
the SC-load voltage matching is not required and the control
algorithm does not require load current sensing. Instead, energy
source current of a much lower amplitude is necessary. The
SC sizing methodology is proposed, and topology issues aiming to
minimize the SC are discussed as well. Compared with a passive
hybrid, the proposed system utilizes much lower capacitance at
the expense of additional power electronics. Experimental results
are presented to demonstrate the feasibility of the approach.
Index Terms DCDC converter, hybrid energy source, power
filter, supercapacitor (SC).

I. I NTRODUCTION

LECTRICAL propulsion loads, as well as power lasers,


pulse modulators employed in plasma applications, and
sensors located in wireless nodes, are characterized by high
peak-to-average power requirements and a low-duty-cycle
operation. In order to maintain sustainable operation of these
systems, the energy source must possess both high-power
and high-energy characteristics. Modern energy sources
mostly possess either of the characteristics, but not both [1].
In case of grid-connected systems and internal combustion
engine-based generators, high power rating means bulky and
expensive connections, and energy conversion devices. As to
portable sources, fuel cells and rechargeable lithium batteries,
characterized by high energy densities, possess poor power
capabilities. Therefore, when designed for powering a pulsed
load, power/energy tradeoffs often result in a nonoptimal
sustainable energy source, providing the required amount
of energy while overpowered and vice versa. One of the

Manuscript received October 25, 2014; revised March 30, 2015; accepted
April 2, 2015. Date of publication April 9, 2015; date of current version
October 29, 2015. Recommended for publication by Associate Editor W. Gao.
The authors are with the Hybrid Energy Sources Research and
Development Laboratory, Department of Electrical Engineering and
Electronics, Ariel University, Ariel 40700, Israel (e-mail: yuhimenko.
vladimir@gmail.com; chaiml@ariel.ac.il; alonku@ariel.ac.il).
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/JESTPE.2015.2421305

feasible solutions is hybridization of high-energy sources with


supercapacitors (SCs) [2], [3]. Recent developments have
demonstrated high charging and discharging power capabilities
of SCs, long shelf/cycle life, and wide-temperature-range
operation, achieved at the expense of low energy density
and relatively high self-discharge. SCs have been recently
employed in a variety of industrial applications [4][12].
Three main ways to hybridize an energy source with an
SC are referred to as passive, semiactive, and fully active
topologies [3], [13]. In the passive topology, which is the most
simple and low cost, the energy source, the SC bank, and the
load are connected in parallel without any power electronics
interface and directly coupled to the load [14], [15]. The
SC employed in such an arrangement is typically oversized
due to a limited terminal voltage swing and the performance
is modest owing to lack of control. On the other hand, the
fully active topology, where both the energy source and the
SC are interfaced to the load by means of dcdc converters,
attains an outstanding performance, compromising the cost,
weight, and volume [16], [17]. The semiactive hybrid, where
either the energy source or the SC are linked to the load
via power electronics interface and the other is connected
to the load directly, presents a tradeoff between performance,
physical dimensions, and price [18], [19]. Some guidelines and
design principles of semiactive hybrids were revealed in [13].
In [20], the battery semiactive hybrid system was developed
and implemented, employing a control approach based on
sensing the load current and forcing the dcdc converter to
supply its average value, thus leaving the SC to supply the
dynamic component of the load demand. Nevertheless, the
approach suffered from two main drawbacks: 1) necessity
of load current sensing and 2) lack of load voltage control.
In order to eliminate both drawbacks, Kuperman et al. [21]
proposed an enhanced capacitance emulating solution, based
on a dual-loop control structure, employing converter rather
than load current sensing. On the other hand, when the energy
source is interfaced by a dcdc converter while the SC is
directly connected to the load (limiting its terminal voltage
swing to load operating voltage range, which is typically
narrow), the low rating of the dcdc converter is offset by the
high SC capacitance, increasing the overall price and possibly
weight and volume of the hybrid system. Therefore, in this
paper, a capacitor semiactive arrangement is studied, where
the SC is interfaced by a dcdc converter while the energy
source is directly connected to the load, leading to reduced
SC capacitance at the expense of a high rating dcdc converter.

2168-6777 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

1002

Fig. 1.

IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 3, NO. 4, DECEMBER 2015

Fig. 2.

Hybridization with the SC.

Fig. 3.

Active power filter. Left: ac. Right: dc.

General pulsed power load.

A dual-loop control cascaded controller is utilized, so that the


overall system resembles the operation of an active power
filter [22], [23], applied to a dc power network, i.e., the
controller forces the energy source to supply the fundamental
(active) component of load power only, letting the dc power
filter to deal with dynamic (reactive) component of the load
power. The SC sizing methodology is discussed for both
analytical and statistical load power profiles, and a tradeoff
between the converter topology and the SC capacitance is
demonstrated.
The rest of this paper is organized as follows. Section II
presents the SC sizing methodology for an arbitrary load
power profile. Converter topology issues are revealed
in Section III, and a design example is given in Section IV.
Small signal analysis and control design of the proposed
system employing a bidirectional buckboost converter
are shown in Section V. Experimental results are given
in Section VI. Finally, the conclusion is drawn in Section VII.

Instantaneous load power can be decomposed into average


and dynamic periodic components as
p L (t) = PL ,AV E + PL ,DY N

N
M


=
Pi,ON Di,ON +
P j,OFF D j,OFF
+

i=1


i=1

II. S IZING M ETHODOLOGY


Consider a pulsed power load possessing a consumption
profile, characterized by periodic multipulse train with
period T and duty cycle D, as shown in Fig. 1. During t DT
interval, the load is in an active ( ON) state, while during
the remaining t (1 D)T interval, the load is in an
idle (OFF) state. Moreover, assume a generalized case in which
statistical distribution rather than exact consumption power
profile of the load is available and given in the histogram
form as

P ,T ,
i = 1, . . . , N
( p) = i,ON i,ON
(1)
P j,OFF , T j,OFF , j = 1, . . . , M
with
Ti,ON = Di,ON T

N


Di,ON = D

i=1
M


T j,OFF = D j,OFF T

D j,OFF = 1 D

(2)

j =1

i.e., the cumulative time of power level Pi appearance is Ti


(i.e., power level Pi may appear more than once, but total
appearance duration is Ti ). In general, each power level can
be positive, zero, or negative (regenerative load).

j =1


2n
t + n .
PLn sin
T

(3)

The zero-average dynamic constituent of the load power


PL ,DYN is equivalent to harmonic + reactive power in
ac distribution systems since the total energy transferred to
the load is average component dependent only
E L = K PL ,AVE T

(4)

with K denoting the number of operating cycles. Nevertheless,


in pulsed power systems, power peak is much higher than the
average component (since the duty cycle D is low), forcing
the energy source to possess significant power density, and
leading to increased rating in case of a single source supply.
Hybridization with an SC (Fig. 2) may minimize the size
and weight of the energy source without compromising the
performance by letting the energy source to supply the average
component of the load only, forcing the SC to struggle with
load dynamics.
Note that this is exactly what shunt active power filters
do in ac distribution systems: connected in parallel to the
load, they supply reactive and harmonic components of load
current, leaving the grid to supply energy-producing component only by forcing grid current to be in phase with the
voltage and resemble its shape [22], as shown in Fig. 3.
Utilizing an SC as dynamic power source is possible since
the net cycle energy of the dynamic component is theoretically

YUHIMENKO et al.: DC ACTIVE POWER FILTER-BASED HYBRID ENERGY SOURCE FOR PULSED POWER LOADS

1003

Equivalent discharge time is then given by


t E Q,ON =

Fig. 4.

Fig. 5.

Et
PMAX,DY N

Pi,ON Ti,ON
.
max Pi,ON

i=1
i

Hence, the SC selection is carried out according to the equivalent discharge profile shown in Fig. 5 as follows. Consider an
SC with capacitance C and equivalent series resistance (ESR)
R in steady-state operation. At an arbitrary instant t = 0 ,
SC terminal voltage is fully charged and resting,
i.e., vSC (0 ) = VSC,MAX . At t = 0+ , constant power
load PMAX,DY N is applied to the device for t E Q,ON period.
As shown in [25], at t = 0+ , internal capacitor voltage
remains unaffected, while terminal voltage changes because
of the ESR voltage drop to



VSC,MAX
4 R PMAX,DY N
+
vSC (0 ) =
. (7)
1+ 1
2
2
VSC,MAX

Dynamic component of general pulsed power load.

From this instant, terminal voltage behaves according to


 
d v1
PMAX,DY N 1
dvSC
SC
+ R PMAX,DY N

vSC = 0
dt
dt
C

Equivalent discharge profile for the SC sizing.

0+ < t < t EQ,ON


zero in both ac and dc systems. This implies that the dynamic
source does not lose energy capacity at the end of operating
cycle (neglecting internal losses). In practice, any powertrain
possesses internal losses that must be compensated by the
energy source.
Following Fig. 1, a dynamic component of the general
pulsed power load is shown in Fig. 4. In case dynamic
power is positive during DT and negative during (1 D)T ,
the SC is aggressively discharged into the load during DT
and then relatively slowly replenished by the energy source
during (1 D)T . In steady state, the operation described
above causes the SC voltage to swing between two extreme
values VSC,MAX and VSC,MIN . While the second-phase
operation is well controlled to prevent the SC voltage from
increasing above maximum allowed terminal voltage (dictated
by the SC chemistry and number of series-connected cells),
in discharge phase, the energy source is unable to cope with
the amount of load power. Consequently, the SC must be
properly selected and sized according to the discharge phase
stress in order to keep the terminal voltage above minimum
allowed value (typically imposed by maximum allowed the
SC powertrain current). According to Fig. 3, the total energy
delivered by the SC during discharge phase is
Et =

N


(5)

i=1

As shown in [24], in case statistical distribution of the load


power profile is given, the design should be accomplished
according to the following worst case: the capacitor should
be sized to be capable of withstanding an equivalent power
burst, delivering E t at PMAX,DY N = max Pi,ON , i = 1, . . . , N
i
appearing during discharge phase, as shown in Fig. 5.

(8)

with initial conditions given by (7). Analytical solution of (8)


is given by the time it takes the terminal voltage to reduce
from vSC (0+ ) to arbitrary voltage v(t) as [24], [25]
t=

C
2PMAX,DY N

 2 +

vSC (0 ) v2 (t) + RC ln

v(t)
.
vSC (0+ )

(9)

Multiplying both sides of (9) by PMAX,DY N energy drawn


from the SC during t E Q,ON interval is obtained as


E 0+ , t EQ,ON = PMAX,DY N t E Q,DY N

C 2 +
vSC (0 ) v2 (t)
2

+ PMAX,DY N RC ln

v(t)
.
vSC (0+ )

(10)

Combining (5) with (10) and substituting v(t) = VSC,MIN ,


the sizing system of equations given by
N

i=1

Pi,ON Ti,ON =


C 2 +
2
vSC (0 ) VSC,MIN
2
+ PMAX,DY N RC ln

Pi,ON Ti,ON .

(6)

VSC,MIN
vSC (0+ )

(11)

and (7) depends on both R and C. If the RC relation is known


or may be estimated [5], it should be used as an additional
equation to explicitly solve for both parameters. Otherwise,
the sizing process must involve multiple iterations to obtain
feasible solution. It should be mentioned that SC parameters
are temperature- and age-dependent; hence, worst case values
should be used in order to ensure a long-term wide temperature
performance sustainability.

1004

Fig. 6.

IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 3, NO. 4, DECEMBER 2015

Capacitor semiactive power filter-based battery SC hybrid topology.

III. T OPOLOGY I SSUES


As for today, power electronics is the least expensive
component of a hybrid power source. On the other hand,
fuel cells, batteries, and SCs are still relatively expensive.
Therefore, in this paper, topology selection is motivated by
the will to reduce the amount of energy sources involved.
As mentioned above, the smallest possible energy source
rating is dictated by average load power plus system losses.
Selecting an energy source according to this criteria imposes
energy requirement (5) on the dynamic power source.
In case an SC is utilized as a power source, a tradeoff
between capacitance and terminal voltage swing exists.
Therefore, in case of a passive hybrid, where both sources are
connected in parallel without any power processing interface,
the SC voltage swing is limited by the energy source terminal
voltage range (typically narrow); hence, relatively high
capacitance is required to meet the specification. In case of an
energy source semiactive hybrid, where the SC is connected
directly across load terminals while the energy source is
interfaced to the load by means of a dcdc converter, the
SC voltage swing is restricted by load voltage requirements,
again leading to increased capacitance requirement. In this
paper, capacitor-semiactive topology shown in Fig. 6 is
selected, in which the energy source is connected directly
across load terminals while the SC is interfaced to the load by
means of a dcdc converter. Once SC is decoupled from the
load, its terminal voltage swing becomes much more flexible,
affected by converter rating and type rather than by load.
Consider a general case with allowed load voltage range
defined by
v L ,MIN < v L < v L ,MAX .

(12)

If the energy source (dc) is realized by means of either


rectified stabilized ac or fuel cell, supplying nearly constant
power would not significantly change its voltage. On the
other hand, in case of battery-based source, its terminal
voltage is expected to vary considerably according to the
state of charge. For example, if the battery pack is formed
by series connection of N LiFePO4 cells, its voltage would
vary between 3.5 N (charged) down to 2.8 N or even
less, i.e., at least 20% terminal voltage decrease is expected
during discharge. Consequently, voltage range limits must be
considered during a design process.

Since an SC energy is approximately proportional to


squared terminal voltage (neglecting internal resistance),
twofold increase of rated voltage theoretically implies fourfold
capacitance decrease. On the other hand, rated voltage increase
requires more cells to be connected in series, alleviating
voltage misbalance issues and increasing ESR. However, these
observations are only valid if the device is cycled from its
rated voltage down to zero, which is obviously impractical
in constant power applications. The majority of previous
works tend to minimize the amount of series-connected cells,
utilizing the SC bank with rated voltage around v R1 = v L ,MIN
in conjunction with bidirectional boost converter. There, the
SC terminal voltage is typically swung from slightly below
v L ,MIN down to 0.5 v L ,MIN . If the SC is required to deliver
an amount of energy given by E, the required capacitance
(again, neglecting internal resistance and conversion losses) is
given by
C1 =

8E
3v2L ,MIN

(13)

If the SC is realized by combining basic cells rated at vRB


with capacitance C B , the required amount of cells would be
N1 =

E
v R1 C1
8
=
.
vRB C B
3 vRB v L ,MIN C B

(14)

On the other hand, if SC with rated voltage of around


v R2 = 1.5 v L ,MAX in conjunction with bidirectional buck
converter is selected while its terminal voltage is swung from
1.5 v L ,MAX down to slightly above v L ,MAX , the required
capacitance reduces to
C2 =

8E
5v2L ,MAX

2
3 v L ,MIN
C1
5 v2L ,MAX

(15)

or C2 < 0.384C1 for the above example at the expense


of 1.5 v L ,MAX /v L ,MIN times increased rated voltage
(1.875 times for the above example). The required amount
of cells is now
E
v R2 C2
12
v L ,MIN

=
= 0.9
N1 (16)
N2 =
vRB C B
5 vRB v L ,MAX C B
v L ,MAX
i.e., 28% less than N1 for the above example. It is important
to note that discussion on the relation between converter
type and Volt-Ampere rating of utilized switches may be found
in [26] while balancing issues have been successfully treated
in [27] and [28]. Consequently, in order to minimize
the required capacitance value, the SC terminal voltage
swing should be increased as possible. To accomplish this,
a buckboost converter is used as the power processing
interface in this paper. If SC with rated voltage of about
v R3 = v R2 = 1.5 v L ,MAX in conjunction with a bidirectional
buckboost converter is utilized while its terminal voltage is
swung from 1.5 v L ,MAX down to 0.5 v L ,MIN , the required
capacitance becomes
C3 =

8E
3
5
C1 =
= 2
C2
2
v
v2L,MIN
v L ,MIN 9 L,MAX 1
9

2
2
v
v

9v2L ,MAX

L,MIN

L,MAX

(17)

YUHIMENKO et al.: DC ACTIVE POWER FILTER-BASED HYBRID ENERGY SOURCE FOR PULSED POWER LOADS

Fig. 7.

1005

Experimental setup of passive battery-SC connection.

or C3 < 0.23C1 and C3 < 0.6C2 for the above example. The
required amount of cells is
E
v R3 C3
4
=
N3 =


v2
vRB C B
3
vRB v L ,MAX 9vL,MIN
CB
L,MAX
=

1
2

v L,MAX
v L,MIN

1
5
N2
v L,MIN N1 =
v2
9vL,MAX
9 v2L,MIN

(18)

L,MAX

or N3 = 0.6N2 = 0.43N1 for the above example. In reality,


the benefit of the third arrangement is slightly lower since
ESR influence should be considered as well.
IV. D ESIGN E XAMPLE
In order to demonstrate the validity of the presented
material, consider a pulsed periodic load, which is given by

640 W, 0 < t < 0.1 T
p L (t) =
(19)
0 W,
0.1 T < t < T
at rated voltage of v L = 12.8 V with T = 5 s. Average
load power and rated current are then 64 W and 5 A,
respectively. Consider a 12.8 V, 5 Ah battery pack, consisting
of 4S1P connection of four automotive LiFePO4 EVIDA
POWER cells (2C maximum continuous discharge rate,
i.e., 10 A) as energy source. Note that while being capable of
supplying the load energy at average rate, the pack is unable to
provide instantaneous load power and hence is hybridized with
SC to endure load power demand. In case the battery supplies
average power only, SC would withstand 576-W power burst
for 0.5 s during discharge phase, delivering 288 J of energy.

Fig. 8.
Experimental results of passive battery-SC connection.
Top: time domain. Bottom: frequency domain.

A. Passive Solution
Considering a low battery ESR (circa 50 m), providing
average current will cause 0.25 V terminal voltage drop.
Consequently, at least 100 F SC is required (in reality, higher
capacitance would be necessary taking ESR into account).
Figs. 7 and 8 demonstrate an experimental setup and results,
respectively, of a passive hybridization between the battery
and two parallel-connected 58 F, 23 m Maxwell SCs.
It may be concluded that the SC act as a high-pass filter
of the load current, as expected from [15]. While the results
are acceptable and battery current is kept below 10 A most
of the time, the SC is obviously oversized and its value
may be significantly reduced using dc active power filter-like
capacitor-semiactive arrangement as shown in the following
subsection.

Fig. 9.

Bidirectional noninverting buckboost converter.

B. Capacitor Semiactive Solution


Consider the power filter-like semiactive topology,
shown in Fig. 6, based on the bidirectional noninverting
buckboost converter, shown in Fig. 9. Converter operation
was recently revealed in [29] and is briefly summarized here.
The converter is shaped by cascaded connection of
two synchronous power legs, sharing common inductor L.
C1 and C2 are high-frequency by-pass capacitors, suppressing
the current ripple and are hence omitted from dynamic analysis
below. Two switches of the same leg are operated in complementary fashion while each leg is driven by a different

1006

IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 3, NO. 4, DECEMBER 2015

Fig. 11.
Fig. 10.

Modulator signals.

Steady-state power-level system.

In frequency domain, (22) and (23) are given as

pulsewidth modulation (PWM) signal. Legs 1 and 2 are


referred to as BUCK and BOOST legs, respectively. The
converter operates as follows. If vSC > v L , the switches of
leg 1 are driven by a PWM signal while Q 2 is OFF and the
converter operates in BUCK mode. If vSC < v L , the switches
of leg 2 are driven by a PWM signal while Q 1 is ON and
the power stage operates in BOOST mode. If vSC v L ,
both legs are PWM-driven and the converter operates in
BUCKBOOST mode. The three mentioned operating modes
are valid for both directions of power flow.
Although converter legs are driven by two different
PWM signals, a single control signal is required to the
modulator in addition to carrier signal vC to create both
converter legs driving PWM signals. Moreover, the modulator
does not need any information regarding either the direction of
the power flow or the operating mode. The input control signal
1 < d < 1 is compared with two similar but shifted carrier
waves vC1 and vC2 to create corresponding PWM driving
signals, as shown in Fig. 10.
As shown in [29], the duty cycles of legs 1 and 2 are
given by

d 1
0,
d1 = K H (1 + d), 1 d V H

1,
d VH

d VL
0,
d2 = K L (d VL ), VL d 1
(20)

1,
d 1
)1 ,

)1 .

K L = (1 VL
respectively, with K H = (1 + V H
Converter operation mode depends on the duty cycle
as follows:

1 d VL
BUCK,
BUCKBOOST, VL d V H
(21)

BOOST,
V H d 1.
In order to act as an active power filter, converter output
current i F should contain all but dc harmonics of the load
current
i F = H P F{i L }

(22)

where the asterisk denotes a desired value and HPF{) is a


high-pass filter operator, leaving the energy source to supply
nearly dc component of the load current
i S (t) = i L (t) i F (t) = i L (t) H P F{i L (t)}.

(23)

i F (s) = H H P F (s) i L (s)

(24)

and
i S (s) = i L (s) i F (s) = i L (s) H H P F (s) i L (s)
= (1 H H P F (s)) i L (s) = H L P F (s) i L (s)

(25)

respectively, with H H P F (s) and H L P F (s) = 1 H H P F (s)


denoting frequency responses of corresponding high- and
low-pass filters. Applying (25) in a single current loop fashion
would theoretically lead to desirable filtering behavior.
However, the approach suffers from two major drawbacks:
1) necessity of load current sensing and 2) instantaneous
power misbalance caused by converter and SC losses,
as shown in Fig. 11, where the steady-state power-level
diagram of the system is depicted. The load power is given
as a sum of average and dynamic components
p L (t) = PL ,AV E + p L ,DY N (t)

576, 0 < t < 0.1 T
= 64 +
[W].
64, 0.1 T < t < T

(26)

The power filter supplies load power dynamic component


while drawing average power (operation losses)
p F (t) = PF,AV E p L ,DY N (t) [W].

(27)

Consequently, the energy source must supply both average


load power component and compensate power filter
losses
pS (t) = PL ,AV E + PF,AV E = 64 + PF,AV E [W].

(28)

Creating energy source current reference according to (25)


will only force it to provide average load power component.
Consequently, power mismatch would occur, forcing the
SC to compensate system losses and deplete eventually.
The second drawback may be eliminated by modifying
(25) as
i S (s) = H L P F (s) (i L (s) + i F (s))

(29)

i.e., considering average power filter current, representing


consumed real power (which is ideally zero). Nevertheless,
this would require additional current sensor. On the other hand,
note that any sourceload average power mismatch directly
affects the SC voltage since power balance is maintained by
drawing/delivering the mismatched power difference from/to
the SC. Subsequently, by regulating the average value of
SC voltage, power balance is maintained and load current

YUHIMENKO et al.: DC ACTIVE POWER FILTER-BASED HYBRID ENERGY SOURCE FOR PULSED POWER LOADS

Fig. 12.

Fig. 13.

Dual-loop cascaded controller structure.

Fig. 14.

Average model of the system.

1007

Feasibility bound of SC sizing.

sensor is no longer required. Note that the dynamic component


of SC voltage must not be influenced by voltage regulation
since it reflects the dynamic component of load power supplied
by the SC. This poses upper limit on voltage regulation
bandwidth.
C. SC Sizing
Consider an SC bank, operating within 157.5 [V]
range. From (26), the energy to be supplied by the
SC during discharge in case of ideal energy converter
is 576 0.1 T = 288 [J]. In order to compensate converter
losses and allow safety margin, 20% overhead is used, leading
to PMAX,DYN = 692 [W] and E t = 346 [J]. Substituting (7) and (11) creates the following sizing equations:



15
4 R 692
+
1+ 1
vSC (0 ) =
2
152

C 2 +
7.5
346 =
v (0 ) 7.52 + 692RC ln
.
2 SC
vSC (0+ )
In case R = 0, resulting capacitance would be Cmin = 4.1 [F].
For R > 0, the required capacitance will obviously increase.
Note that in order to be properly sized, selected capacitor
values must reside above the curve bound shown in Fig. 12,
i.e., in case of low quality high ESR capacitors, SC value
may be very high. In addition, capacitors with R > 81 m
are nonfeasible since the expression inside the square root
would be negative, indicating that initial voltage drop is
higher than 7.5 [V]. It is interesting to point out that in case
of R 75 m, the capacitance would approach that of passive
solution. Consequently, the SC selection is nontrivial and must
be accomplished carefully.
It should be emphasized that the selected capacitor
parameters should reside as close to feasibility bound as
possible in order to maximally utilize the SC.
V. S MALL S IGNAL A NALYSIS AND C ONTROL D ESIGN
Cascaded voltagecurrent dual-loop controller is proposed,
as shown in Fig. 13. The slow outer loop ensures the average
power balance by comparing the SC voltage to a desired
reference value. The outer loop controller C V (s) calculates the
desired energy source current i S essential for maintaining the

power balance. The fast inner loop controller C I (s) determines


the necessary duty cycle for the converter to output current to
follow iL i S thus forcing energy source current to follow i S .
The inner loop plant PI (s) represents the transfer function
from the duty cycle to the energy source current, while the
output loop plant PO (s) denotes the transfer function from
the energy source current to the SC voltage.
A. Current Loop
In order to derive the inner loop plant, necessary for current
controller design, consider averaged model of the system,
shown in Fig. 14, where the energy source is represented by its
Thevenin equivalent (VS , R S ). Both Thevenin parameters vary
slow enough to be considered constant within the bandwidth
of interest. The set of equations, describing the system, is then
given by
L

di I
+ R L i I = d1 v L (1 d2 )vSC
dt
dvC
= (1 d2 )i I
C
dt
dvC
vSC = vC + RC
dt
pL
i S = d1 i I +
vL
v L = VS R S i S

(30)

with R L denoting inductor series resistance. A fast current


loop is designed in this subsection, allowing inner and outer
loops decoupling (loop bandwidths should differ by order of
magnitude or higher). Inner loop related part of (30) may be
rearranged as
L

di I
= d1 VS (1 d2 )vSC d1 R S i S R L i I
dt
i S = d1i I + VS1 R S i S (i S d1i I ) + VS1 p L .

(31)

Splitting the time-based variables into dc and ac components


d1 = D1 + d1 , d2 = D2 + d2 , i S = I S + iS
i I = I1 + iI , p L = PL + p L

(32)

1008

IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 3, NO. 4, DECEMBER 2015

and substituting into (31) the small-signal dynamics is


obtained as
Ls iI = d1 VS + d2 VSC (1 D2 )vSC d1 R S I S iS R S D1
R L iI = d1 VS + d2 VSC f1
i S = d1 I I + iI D1 + V 1 (iS R S (2I S D1 I I ) d1 I I R S I S
S
iI D1 R S I S ) + VS1 p L
= d1 I I + iI D1 + f2 + VS1 p L
(33)
linearized around an equilibrium point given by
D1 (VS R S I S ) (1 D2 )VSC r L I L = 0
VS
PL
I S2
IS +
= 0.
(34)
RS
RS
Note that the ESR values of inductor, battery, and SC are
difficult to obtain in practice. Moreover, they can vary
significantly during operation. Consequently, all ESR-related
terms in (33) are lumped into uncertainty terms f 1 and f 2 .
Further rearranging reveals the small-signal energy source
current dynamics


i S = L I I s + D1 VS d1 + D1 VSC d2 + f2 f1 D1 + V 1 p L
S
Ls
Ls
Ls
D1 VSC
L I I s + D1 V S
(35)
=
d1 +
d2 + fi
Ls
Ls
where f i is the total uncertainty and disturbance of the current
loop. In addition, according to (20) the following holds:

d 1
d VL
0,
0,

d1 = K H d, 1 d V H d2 = K L d, VL d 1

0,
d VH
0,
d 1

d 1
0,
D1 = K H (1 + D), 1 d V H

1,
d VH

d VL
0,
(36)
D2 = K L (D VL ), VL d 1

1,
d 1.

Fig. 15.

Low-frequency power-level system.

is limited by the worst case of the lowest unstable


zero frequency (equivalent to highest value of ), occurring
in BUCK mode. As a rule of thumb, current loop bandwidth
may be chosen as one-fifth of the lowest unstable
zero frequency
Ci L =

1 K H (1 + DMIN )VS,MIN
.
5
L|I I,MAX |

(39)

B. Voltage Loop
Once current loop is properly closed, allowing to assume
that inner and outer loops are decoupled, outer loop
controller may be designed based on average power balance
principle [30] as follows. Consider low-frequency power-level
system diagram shown in Fig. 15, where dynamic components
of load and power filter power is omitted since they reside
outside the bandwidth of interest. Note that the power
converter losses PL ,AV E are included in the model as well.
The system in Fig. 15 is described by


dvC
= PF,AV E + PL ,AVE
PS + PC = VS R S i S i S vSC C
dt
dvC
(40)
vSC = vC + RC
dt
where PC and vC denote power, flowing out of SC and internal
SC voltage, respectively. Rearranging (40) as
 2
dvC
= PF,AVE + PL ,AVE + R S i S  PAVE
VS i S vSC C
dt
dvC
(41)
vSC = vC + RC
dt
and applying small-signal analysis results in

VS 1 + RCs 
v SC =
(42)
i S VS1 PAVE .
VSC C
s
Recall that voltage loop bandwidth is limited by the lowest
component of dynamic load power spectrum. As a rule
of thumb, the voltage loop bandwidth may be chosen as
[with respect to (19)] one-tenth of fundamental load power
frequency

Hence, small-signal energy source current dynamics in each


of the operation modes may be expressed by combining (35)
with (36) as
1 + s
iS = G
(37)
d(s) + fi (s)
s
with
2 (1 + D)V
KH
S

K H (1+ D) (K H VS + K L VSC )
G=

K
(1
+
D)V
K
.
(43)
Cv L =
L
H
SC

10 T
L

L II

VI. VALIDATION

,
BUCK

K H (1 + D)VS
Experimental setup of a capacitor semiactive hybrid
L II
(38) is shown in Fig. 16, consisting of 12.8 V/5 Ah battery
=
, BUCKBOOST

(1 + D) (K H VS + K L VSC )

pack, formed by 4S1P connection of four automotive

0,
BOOST.
LiFePO4 EVIDA POWER cells (2C maximum continuous
According to (38), the dynamics is nonminimum phase for discharge rate, i.e., 10 A), 5.5 F/7.5 m SC bank
I I < 0 (i.e., SC supplies power to the load) in BUCK and (satisfying Fig. 12) and custom buckboost converter
BUCKBOOST modes. Consequently, current loop bandwidth (35-H
inductor,
20-kHz
switching
frequency),

YUHIMENKO et al.: DC ACTIVE POWER FILTER-BASED HYBRID ENERGY SOURCE FOR PULSED POWER LOADS

Fig. 16.

1009

Experimental setup.
Fig. 18.

Experimental battery and SC voltages of active battery-SC hybrid.

to Fig. 12, for R = 7.5 m capacitance of 4.6 F would be


sufficient. Nevertheless, forming such an SC bank was
practically impossible due to the commercially available
SC sizes.
VII. C ONCLUSION
Design and implementation of a capacitor semiactive
hybrid energy source based on dc power filter principle
was described in this paper, including presenting the
sizing methodology and topology issues. The system under
study contains an SC interfacing bidirectional buckboost
dcdc converter, supporting an energy source connected to a
pulsed power load. The control algorithm was adopted from
the power filter theory, aiming to reduce the current of the
source despite operating under high peak-to-average power
dc load by forcing the SC to supply the dynamic component
of the load power. Moreover, energy source current only
is required in the controller, allowing to utilize a relatively
low rating sensor. It was shown that compared with a
passive hybrid, the proposed system utilizes 20 times lower
capacitance at the expense of additional power converter.
Experimental results were given to validate the viability of
presented findings.
R EFERENCES
Fig. 17. Experimental currents of active battery-SC hybrid. Top: time domain.
Bottom: frequency domain.

operated by an ATMEGA microcontroller. Voltage and


current loops were compensated using ProportionalIntegrative
controllers
with
bandwidths
chosen
as 0.02 Hz (= 0.1/T) and 1 kHz, respectively.
Experimental results are shown in Figs. 17 and 18.
Semiactive hybrid performance is quite similar to time and
frequency domain results of passive hybrid. Battery current is
kept below 10 A at all time as requested while dc power filter
provides the majority of dynamic current. Moreover, note that
the SC voltage does not reach the lower bound of 7.5 [V] since
its parameters reside above the feasibility bound. According

[1] R. M. Dell and D. A. J. Rand, Energy storageA key technology


for global energy sustainability, J. Power Sour., vol. 100, nos. 12,
pp. 217, Nov. 2001.
[2] A. Burke, Ultracapacitor technologies and application in hybrid and
electric vehicles, Int. J. Energy Res., vol. 34, no. 2, pp. 133151, 2009.
[3] A. Kuperman and I. Aharon, Batteryultracapacitor hybrids for pulsed
current loads: A review, Renew. Sustain. Energy Rev., vol. 15, no. 2,
pp. 981992, 2011.
[4] J. M. Blanes, R. Gutierrez, A. Garrigos, J. L. Lizan, and J. M. Cuadrado,
Electric vehicle battery life extension using ultracapacitors and an
FPGA controlled interleaved buckboost converter, IEEE Trans. Power
Electron., vol. 28, no. 12, pp. 59405948, Dec. 2013.
[5] O. Laldin, M. Moshirvaziri, and O. Trescases, Predictive algorithm for
optimizing power flow in hybrid ultracapacitor/battery storage systems
for light electric vehicles, IEEE Trans. Power Electron., vol. 28, no. 8,
pp. 38823895, Aug. 2013.
[6] J. Cao and A. Emadi, A new battery/ultracapacitor hybrid energy
storage system for electric, hybrid, and plug-in hybrid electric vehicles,
IEEE Trans. Power Electron., vol. 27, no. 1, pp. 122132, Jan. 2012.

1010

IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 3, NO. 4, DECEMBER 2015

[7] D. Iannuzzi and P. Tricoli, Speed-based state-of-charge tracking control


for metro trains with onboard supercapacitors, IEEE Trans. Power
Electron., vol. 27, no. 4, pp. 21292140, Apr. 2012.
[8] S.-M. Kim and S.-K. Sul, Control of rubber tyred gantry crane with
energy storage based on supercapacitor bank, IEEE Trans. Power
Electron., vol. 21, no. 5, pp. 14201427, Sep. 2006.
[9] H. Yang and Y. Zhang, Analysis of supercapacitor energy loss for power
management in environmentally powered wireless sensor nodes, IEEE
Trans. Power Electron., vol. 28, no. 11, pp. 53915403, Nov. 2013.
[10] F. Ongaro, S. Saggini, and P. Mattavelli, Li-ion battery-supercapacitor
hybrid storage system for a long lifetime, photovoltaic-based wireless sensor network, IEEE Trans. Power Electron., vol. 27, no. 9,
pp. 39443952, Sep. 2012.
[11] S. Kim and P. H. Chou, Size and topology optimization for
supercapacitor-based sub-watt energy harvesters, IEEE Trans. Power
Electron., vol. 28, no. 4, pp. 20682080, Apr. 2013.
[12] B. Hredzak, V. G. Agelidis, and M. Jang, A model predictive control system for a hybrid battery-ultracapacitor power source, IEEE
Trans. Power Electron., vol. 29, no. 3, pp. 14691479, Mar. 2014.
[13] I. Aharon and A. Kuperman, Topological overview of powertrains for
battery-powered vehicles with range extenders, IEEE Trans. Power
Electron., vol. 26, no. 3, pp. 868876, Mar. 2011.
[14] R. A. Dougal, S. Liu, and R. E. White, Power and life extension of
battery-ultracapacitor hybrids, IEEE Trans. Compon. Packag. Technol.,
vol. 25, no. 1, pp. 120131, Mar. 2002.
[15] A. Kuperman, I. Aharon, A. Kara, and S. Malki, A frequency domain
approach to analyzing passive batteryultracapacitor hybrids supplying
periodic pulsed current loads, Energy Convers. Manage., vol. 52, no. 12,
pp. 34333438, 2011.
[16] M. B. Camara, H. Gualous, F. Gustin, and A. Berthon, Design and new
control of DC/DC converters to share energy between supercapacitors
and batteries in hybrid vehicles, IEEE Trans. Veh. Technol., vol. 57,
no. 5, pp. 27212735, Sep. 2008.
[17] Z. Amjadi and S. S. Williamson, Power-electronics-based solutions for
plug-in hybrid electric vehicle energy storage and management systems,
IEEE Trans. Ind. Electron., vol. 57, no. 2, pp. 608616, Feb. 2010.
[18] J. Moreno, M. E. Ortuzar, and J. W. Dixon, Energy-management system
for a hybrid electric vehicle, using ultracapacitors and neural networks,
IEEE Trans. Ind. Electron., vol. 53, no. 2, pp. 614623, Apr. 2006.
[19] M. Ortuzar, J. Moreno, and J. Dixon, Ultracapacitor-based auxiliary
energy system for an electric vehicle: Implementation and evaluation,
IEEE Trans. Ind. Electron., vol. 54, no. 4, pp. 21472156, Aug. 2007.
[20] L. Gao, R. A. Dougal, and S. Liu, Power enhancement of an actively
controlled battery/ultracapacitor hybrid, IEEE Trans. Power Electron.,
vol. 20, no. 1, pp. 236243, Jan. 2005.
[21] A. Kuperman, I. Aharon, S. Malki, and A. Kara, Design of a semiactive battery-ultracapacitor hybrid energy source, IEEE Trans. Power
Electron., vol. 28, no. 2, pp. 806815, Feb. 2013.

[22] P. Acuna, L. Moran, M. Rivera, J. Dixon, and J. Rodriguez, Improved


active power filter performance for renewable power generation systems, IEEE Trans. Power Electron., vol. 29, no. 2, pp. 687694,
Feb. 2014.
[23] W.-H. Choi, C.-S. Lam, M.-C. Wong, and Y.-D. Han, Analysis of
DC-link voltage controls in three-phase four-wire hybrid active power
filters, IEEE Trans. Power Electron., vol. 28, no. 5, pp. 21802191,
May 2013.
[24] A. Kuperman et al., Supercapacitor sizing based on desired power and
energy performance, IEEE Trans. Power Electron., vol. 29, no. 10,
pp. 53995405, Oct. 2014.
[25] M. Mellincovsky et al., Performance and limitations of a constant
power-fed supercapacitor, IEEE Trans. Energy Convers., vol. 29, no. 2,
pp. 445452, Jun. 2014.
[26] G. Guidi, T. M. Undeland, and Y. Hori, An interface converter
with reduced VA ratings for battery-supercapacitor mixed systems, in
Proc. IEEE Power Convers. Conf., Nagoya, Japan, Apr. 2007, pp. 936
941.
[27] Y. Yuanmao, K. W. E. Cheng, and Y. P. B. Yeung, Zero-current switching switched-capacitor zero-voltage-gap automatic equalization system
for series battery string, IEEE Trans. Power Electron., vol. 27, no. 7,
pp. 32343242, Jul. 2012.
[28] M. Uno and K. Tanaka, Single-switch multioutput charger using voltage
multiplier for series-connected lithium-ion battery/supercapacitor equalization, IEEE Trans. Ind. Electron., vol. 60, no. 8, pp. 32273239,
Aug. 2013.
[29] I. Aharon, A. Kuperman, and D. Shmilovitz, Analysis of dual-carrier
modulator for bidirectional noninverting buckboost converter, IEEE
Trans. Power Electron., vol. 30, no. 2, pp. 840848, Feb. 2015.
[30] S. Singer and R. W. Erickson, Power-source element and its properties, IEE Proc.-Circuits, Devices Syst., vol. 141, no. 3, pp. 220226,
Jun. 1994.

Vladimir Yuhimenko, photograph and biography not available at the time of


publication.

Chaim Lerman, photograph and biography not available at the time of


publication.

Alon Kuperman, photograph and biography not available at the time of


publication.

Você também pode gostar