Evaluation of Shunt and Series Power Conditioning
Strategies for Feeding Sensitive Loads
Bingsen Wang
Giri Venkataramanan
Department of Electrical and Computer Engineering
University of Wisconsin - Madison
Madison, WI 53706-1607
Email: bingsen@cae.wisc.edu
Abstract— Proliferation of electronic equipment in commercial
and industrial processes has resulted in increasingly sensitive
electrical loads to be fed from power distribution systems. In
this paper, series and shunt compensation schemes for feeding
sensitive load centers are studied. The purpose of the system-level
study is to reveal insights about these two mitigation strategies so
that designers may select an appropriate compensation strategy
for a given application. The paper presents analytical models
for comparing series and shunt power conditioning approaches
on the basis of power converter rating. Operating strategies,
computer simulations and experimental verification of a series
power conditioning approach are presented.
I. I NTRODUCTION
In the recent past, dramatic improvements in productivity
have been realized in the high technology sector as well as
in the traditional industrial sector. From a point of view of
the electric power supply to these segments of loads, this
increase in productivity has led to a concomitant increase in
the number of loads that are sensitive to power quality. Some
of the industries that have such large sensitive loads include
semiconductor manufacturing, textile mills, paper mills and
plastic injection molding, etc. Of course, a number of smaller,
but equally critical loads such as computers and electronic data
processing equipment are also sensitive to power quality. The
tolerance levels of computer equipment are specified by the
ITI/CBEMA (Information Technology Industry/Computer &
Business Equipment Manufacturers’ Association) curves [1].
Similar performance specifications for voltage quality have
been developed for semiconductor manufacturing industries
as well. Fig. 1 illustrates such voltage sag susceptibility
curves. These curves plot the percent of nominal voltage at
the load terminals as a function of duration in cycles of
the ac supply (60 Hz). They represent the susceptibility of
systems to voltage sags that are repetitive and intermittent. The
CBEMA curves represent the boundary of the ac input voltage
envelope that can be tolerated (typically) by most computerbased equipment. When the supply voltage is between the
upper curve and the lower curve, the equipment will continue
to function normally. When the supply voltage falls outside
this boundary, the equipment typically stops functioning.
As seen in Fig. 1, the steady state range of tolerance for
sensitive equipment is 10% from the nominal voltage, i.e.,
the equipment continues to operate normally when sourced by
any voltages in this range for an indefinite period. Similarly,
0-7803-8269-2/04/$17.00 (C) 2004 IEEE
Fig. 1. Curves that indicate acceptable power quality power levels of sensitive
equipment.
voltage swells to magnitude 120% of the nominal value can be
tolerated for about 0.5 s or 30 cycles; voltage sags to 80% of
nominal for 10 s, i.e., 600 cycles can be tolerated [2]. When the
supply voltage is outside the boundaries of the susceptibility
curves, improvement of quality of power supplied to sensitive
loads is essential in order to avoid a possible interruption in
their operation. The most straightforward way of ensuring
adequate power quality levels is the use of uninterruptible
power supply (UPS) systems. However, as the size of plant
increases, this solution becomes uneconomical. Alternative
solutions to the problem stem from a careful study of the
nature of power quality problems in the field.
Various field power quality surveys indicate that the majority of events that result in process disruption are caused
by voltage sags. Moreover, over 90% of voltage disturbances
in the utility lines are single-phase voltage sags caused by
momentary single-line to ground (SLG) faults in distribution
systems [3]. The short circuit faults can be caused by lightning strokes, storms, animals and other unpredictable factors.
Therefore, improvement of power quality levels should begin
with reduction of voltage sags, particularly in a per phase
basis. Although the voltage sag levels can be reduced at
the power system level by oversizing the components of the
1445
distribution system installations, often such solutions are too
costly to implement. Alternatively, utilities and customers turn
to the practical solution of feeding premium power to sensitive
loads, which is realized by the mitigation equipment at the
distribution feeder close to the customer. This solution is also
called Custom Power (CP) [4].
Among CP technologies, both series and shunt devices
have been applied to solve the voltage sag/dip and flicker
problems and some of these devices are commercialized [5],
[6], [7]. Due to the project-oriented nature of CP design, few
general rules can be found in literature. Although extensive
research work on the device, either series or shunt, has been
carried out [8], [9], most of them are focused on the device,
not at the system level. In this paper, the series and shunt
compensation schemes are studied at the system level. The
purpose of this system-level study is to reveal more insights
about these two compensation strategies so that designers of
the CP device would have some guidance when selecting
compensating strategies at the very beginning of a project.
Another motivation of this work is to lay a ground for the
application of H-bridge multilevel convert to dynamic voltage
restorer (DVR). Different operation strategies are evaluated
and compared regarding the rating issues, operation and fault
modes.
At the outset, one of the first steps in evaluating a particular
power conditioning approach is to determine the VA rating of
the power converter illustrated as a voltage source converter
(VSC) in Fig. 2. Rating of the power converter depends
on the load to be serviced, the depth of voltage sag to be
compensated and the properties of the power system at the load
location. The dependencies among these quantities for the two
compensation approaches are developed in Section II. Once
the power conditioning system ratings have been determined,
their operation during sags can be steered such that there is
no active power supplied by the power conditioner, when the
operating conditions permit. These operating boundaries are
determined and presented in Section III. Furthermore, when
the operating conditions demand that active power need to
be supplied by the power conditioner, it would be desirable
to draw the least possible amount of active power from
the storage device to maximize the ride-through capability.
Operation under these conditions is presented in Section IV.
Computer simulation results under this mode of operation
are presented in Section V. Experiment with 7-level H-bridge
cascaded multilevel converter has been carried out and experimental results are included in Section VI. Transitions between
various operating modes, faulted and bypassed conditions are
discussed in the concluding section.
Fig. 2. Schematic representation of commonly used mitigation approaches
to improve voltage sag immunity of sensitive loads: (a) series type; (b) shunt
type
load as an impedance ZL , with a power factor angle φ. For
the purpose of sizing the VA rating of the power conditioner,
the series device is modelled as a voltage source Vinj and the
shunt device is modelled as a current source Iinj .
For a general load with a wide variation in power factor
to be fed under deep sag conditions, the power conditioner
typically has to supply active power and reactive power. The
size of the power conditioner, including interface transformers
if any, is determined by the apparent power injection. Without
any loss of generality, the following assumptions have been
made to size the power conditioner used as the compensation
device:
•
II. P OWER C ONVERTER R ATING - M INIMUM A PPARENT
P OWER
•
The simplified single-line equivalent circuits for the series
and shunt compensation approaches are illustrated in Fig. 3(a)
and Fig. 4(a), respectively. The source is modelled by its
Thevenin equivalent voltage source Vs and series reactance
Xs seen at the point of common coupling (PCC), and the
The magnitude of the load current is 1 per unit. Its phase
angle is referenced to the load voltage.
The magnitude of the compensated load voltage is maintained at 1 per unit.
With these assumptions, the problem of minimizing apparent
power turns to minimizing the injected voltage for series
compensation while minimizing the injected current for shunt
compensation.
1446
Fig. 3. Illustration of series compensation approach to improving power
quality: (a) single-line equivalent circuit representing key parameters; (b)
phasor diagram of salient quantities
Fig. 4. Illustration of shunt compensation approach to improving power
quality: (a) single-line equivalent circuit representing key parameters; (b)
phasor diagram of salient quantities
A. Series Compensation Strategy
For a series compensation approach, the following phasor
equations are readily written from circuit representation as
shown in Fig. 3(a).
Vs
=
V pcc + jI L Xs
VL
=
V pcc + V inj
VL
=
I L |ZL |e
(1)
jφ
where the under-bars (“ ”) with voltage and current variables
indicate they are phasor quantities. In the following text,
current and voltage variables without under-bars represent
corresponding magnitude of those phasors. The phasor diagram illustrating the relationship expressed by (1) is shown in
Fig. 3(b).
In the phasor diagram, the load current is conveniently
chosen as the reference phasor for the series circuit and the line
voltage is represented by the phasor OA, leading the current
by a phase angle φ. It is clear from the phasor diagram, that
the locus of the voltage at PCC is an arc (with radius = Vs )
centered at O , which is vertically displaced from the center
of the locus arc of vector Vs by IL Xs . In this case, the phasor
of the injected voltage is collinear with the segment O A,
and represents its minimum value. Other values of injected
voltage values may be possible while maintaining 1 p.u. output
voltage, but will not represent a minimum value of injected
voltage. Using geometric relationships, the magnitude of the
injected voltage under this condition can be expressed as
Vinjmin = VL2 + (IL Xs )2 + 2VL IL Xs sin φ − Vs . (2)
It may be noted that the magnitude of the minimum injected
voltage increases as the Thevenin impedance of the system
increases.
B. Shunt Compensation Strategy
In a similar manner, for the shunt compensation approach,
the following phasor relations will hold from Fig. 4(a).
Vs
IL
= V pcc + jI L Xs
= I s + I inj
VL
= I L |ZL |e
(3)
jφ
The phasor diagram illustrating the above relations is shown
in Fig. 4(b).
In the phasor diagram, the load voltage is conveniently
chosen as the reference phasor for the parallel circuit and
the load current is represented by the phasor OA, lagging
the voltage by a phase angle φ. The locus of the source
current is the arc (radius = Vs /Xs ) centered at O , which is
vertically displaced from the center of the locus arc of vector
Vs by VL /Xs . The minimum injected current lines up with
the segment O A. In this case, the injected current can be
expressed as
VL
2VL IL
Vs
Iinjmin = ( )2 + IL2 +
sin φ −
.
(4)
Xs
Xs
Xs
It may be noted that the magnitude of the minimum injected
current decreases as the Thevenin impedance of the system
increases.
1447
90
Shunt Injection with Vs=0.8
SCR= 8
SCR= 4
SCR= 2
SCR= 1
60
90
Series Injection with Vs=0.8
30
60
30
φ
φ
0.5
1
VA rating (pu)
0
1.5
(a)
0.5
1
1.5
VA rating (pu)
2
2.5
3
0
(b)
Fig. 5. Power converter VA ratings for 20% sag for (a) series (b) shunt
compensation approaches. (Radius corresponds to the VA rating of the
converter and the angle of corresponds to the power factor angle of the load
respectively.)
Fig. 6. Variation of power converter VA ratings as a function of power factor
and (sagged) source voltage with SCR=8 for a series injection strategy.
C. Comparative Ratings
Based upon the above relationships (2) and (4), the ratings
of the power conditioner may be illustrated in the form of polar
plots as shown in Fig. 5, with radius and angle coordinates.
In these plots, the radius represents the VA rating of the
power conditioner and the angle represents the lagging (as is
commonly the case in an industrial power system) power factor
angle of the load current while correcting a sag of 20% depth
below the nominal voltage. As may be observed from (2) and
(4), the VA ratings depend on the system impedance and hence
the short-circuit ratio (SCR) at PCC. This behavior is readily
evident in the various curves plotted in Fig. 5, for different
values of SCR ranging from 1 to 8. It may be observed from
the figure that the VA rating of the shunt device is greater than
the series counterpart for the case of SCR larger than unity. In
practical distribution system, the SCR is generally greater than
the cases shown in the plots. So the converter sizing favors
series compensation approach if the voltage correction is the
main goal of the application. Therefore, further discussions
are primarily focused on the series compensation strategy,
although aspects of the shunt compensation approach will be
pointed when appropriate. In the case of a series compensation
approach for a given system (with fixed SCR), the VA rating
depends on the load power factor and voltage sag depth. The
relation is illustrated in Fig. 6 for series injection strategy with
the power rating represented by the surface.
III. Z ERO ACTIVE P OWER O PERATION S TRATEGY
Once power converter ratings have been fixed on the basis
of operating power factor and desired depth of sag to be
corrected, various operating modes may be used to mitigate
typical voltage sag conditions. If the load is not sensitive to the
voltage phase jump, it is possible to correct voltage sag with
purely reactive power compensation (RPC). Although RPC
Fig. 7. Phasor diagram of salient quantities for series compensation under
RPC mode of operation.
mode of operation is limited in terms of voltage sag depth,
it does not deplete the power conditioner energy storage, and
hence can provide ride through for an infinite period of time.
To be sure, the injected values of voltages will have to be
within the design limits of the converter chosen on the basis of
the discussion in the previous section. The phasor diagram of
various quantities of the system under RPC mode of operation
are illustrated in Fig. 7. In this case the injected voltage leads
the load current by 90 degrees. In contrast, the injected current
lags the load terminal voltage by 90 degrees for the shunt
compensation approach.
The limiting case for series compensation to operate under
RPC mode may be determined to be
Vs(min
RP C)
= VL cos φ.
(5)
Furthermore, at the operating condition, the required voltage
1448
90
Series Injection with Vs=0.8
SCR=1
SCR=2
SCR=4
SCR=8
60
30
φ
(P)
0.5
1
VA rating (pu)
1.5
2
0
Fig. 8. Power converter VA injection for 20% sag under RPC mode of
operation. (Radius corresponds to the VA rating of the converter and the
angle of corresponds to the power factor angle of the load respectively.)
Fig. 9. Phasor diagram of salient quantities for series compensation under
minimum active power mode of operation.
90
can be found to be
Vinj = VL sin φ −
Series Injection with Vs=0.8
SCR=1
SCR=2
SCR=4
SCR=8
60
Vs2 − (VL cos φ)2 + IL Xs .
(6)
The converter VA injection of series compensation is plotted
in Fig. 8 for 20% voltage sag. The polar plots in the figure
show the variation of converter VA ratings for different SCRs,
as the load power factor angle varies. It is clear that the
operation region under zero active power compensation mode
will be limited by the load power factor. Of course, it will be
also limited by available device ratings.
30
φ
0.5
IV. M INIMUM ACTIVE P OWER O PERATION S TRATEGY
When the operation conditions are beyond the limits of pure
reactive power compensation, a minimum active power (MAP)
injection strategy may be used so that the energy storage
elements can be minimized, or the depletion rate of stored
energy is minimal for servicing the load at a given operating
condition. In particular, the source voltage falls below the
limits given by (6), the RPC strategy fails, and one needs
to resort to active power injection.
To minimize the active power injection from the power
conditioner, active power extraction from the source Vs should
be maximized. The phasor diagram of the salient quantities of
the system under this mode operation is shown in Fig. 9. For
series compensation, the active power supplied by the source
is maximized under unity power factor operation, i.e. source
voltage is in phase with load current. This minimizes the power
to be supplied from the VSC. This is also clear from the phasor
diagram that Vinj has minimum projection on IL . In this case,
the injected voltage becomes
Vinj = (VL cos φ − Vs )2 + (VL sin φ + IL Xs )2 .
(7)
The converter VA injection required for 20% sag is plotted
in Fig. 10. The MAP strategy can be considered to be an
extension of RPC. When the converter operating conditions
fall outside the operating region possible with RPC or MAP,
1
VA rating (pu)
1.5
2
0
Fig. 10. Power converter VA injection with minimum active power operation
strategy for 20% voltage sag. (Radius and angle of the plot correspond to VA
rating of the converter and power factor angle of the load, respectively.)
the converter is operated in the minimum apparent power (or
baseline) operating mode.
V. C OMPUTER S IMULATION R ESULTS
In order to verify the operation of the system according
to the analytical models, a computer simulation has been
carried out for a series power conditioning device using
Matlab and Simulink [10], [11]. The block diagram of the
controller is shown in Fig. 11. Selected results from the system
operating under RPC mode are shown here. The structure of
the controller and its design are beyond the scope of the current
paper and are discussed elsewhere [12].
The parameters of the system used in the simulation are:
√
Ls = 1.4 mH, L = 10 mH, R = 2 Ω, VAC = 480/ 3
V(rms). These parameter settings correspond to SCR = 8.0,
load power factor= 0.47. The simulation results with a 20%
voltage sag in three phases are illustrated in Fig. 12. The figure
illustrates the source voltage, injected voltage, load voltage,
load current and inverter dc bus voltage. After transient, only
1449
Source voltage
Injection voltage
−500
−200
0.5
0
5
10
15
a
b
c
0
−500
0
5
10
15
100
a
b
c
0
0
5
10
15
DC bus
data 1
700
680
0
5
10
15
Source voltage
Fig. 12. Computer simulation results of the series compensation approach
under a three-phase balanced sag (the plots are source voltages, series injection
voltages, load terminal voltages, load currents and device internal dc bus
voltage, respectively, from top to bottom.)
500
a
b
c
0
−500
0
5
10
15
200
a
b
c
0
−200
0
5
10
15
500
a
b
c
0
−500
0
5
10
15
100
Load current
TABLE I
PARAMETERS USED IN EXPERIMENT.
Load power factor
a
b
c
t
Injection voltage
A laboratory prototype voltage conditioning system has
been developed using a H-bridge cascaded multilevel converter. The converter has been used to verify the predictions
from the analytical modelling during operation in RPC mode.
The parameters of the laboratory prototype system (in pu),
with a base power of 35 VA and a base voltage of 25 V, are
listed in Table I.
1.0
15
500
660
Load voltage
VI. E XPERIMENTAL R ESULTS
Load impedance ZL
10
720
reactive power is injected, which can be seen from the steady
state settlement of dc bus voltage, which is purely capacitive.
The injection voltage is about 103 V(rms), or 0.37 pu. This
result is indicated as point ’P’ in Fig. 8. It is clear this result
matches the prediction. Although the analytical model and the
controller have been developed only for a balanced three phase
system, the performance of the system for a more typical
system with unbalanced sag may be studied readily using
computer simulations. The response of the system With the
same controller under a single phase voltage sag is shown in
Fig. 13.
0.115
5
0
Controller block diagram
Source impedance Xs
0
200
−100
Fig. 11.
a
b
c
0
Load current
Load voltage
500
a
b
c
0
−100
0
5
10
15
Selected traces from the experimental system are shown
in Fig. 14. The injected voltage (the 2nd trace) leads the
load current (the 4th trace) by 90 degrees, indicating that
only reactive power is being injected from the VSC. The
source voltage Vs = 0.115 with the load voltage and load
current both are one per unit. With the measured quantities and
parameters given in Table I, the series injection voltage can be
DC bus
710
vdc
700
690
680
0
5
10
15
t
Fig. 13. Computer simulation results of the series compensation approach
under a single-phase voltage sag (the plots are in the same order as in Fig. 12).
1450
Other operating issues such as storage system management,
fault management, bypass, mode transitions, controller design
issues etc. are current issues being studied further and the
results from these studies will be presented in future publications.
ACKNOWLEDGMENT
The authors would like to acknowledge support from Wisconsin Electric Machine and Power Electronics Consortium
(WEMPEC) at the University of Wisconsin-Madison. The
work made use of ERC shared facilities supported by the
National Science Foundation (NSF) under AWARD EEC9731677. The laboratory prototype system used for the experimental work was developed by Mr. Yusuke Fukuta through
support from the Office of Naval Research through grant
N00014-01-1-0623.
R EFERENCES
Fig. 14. Traces of source voltage, series injected voltage, load terminal
voltage and load current, respectively, from top to bottom obtained from the
laboratory prototype system.
calculated as 0.741 pu by use of equation (6). The measured
injection voltage is 0.752 pu. This small error between the
analytical predictions and experimental measurements (about
1.5%) may be attributed to parameter uncertainties and losses
in the system.
VII. C ONCLUSIONS
The application power quality conditioning devices to meet
stringent demands of sensitive loads is continuing to grow.
Although series and shunt compensation approaches for providing power quality conditioning have been proposed and
demonstrated, clear association between operating scenarios
and their characteristics have not been definitively known. This
paper has presented a simple but useful analytical modelling
approach to quantify the the power converter ratings required
for the two approaches. On the basis of the modelling results,
it may be concluded that series power conditioning approach
is suitable for strong ac systems and shunt power conditioning
approach is suitable for weak ac systems.
Based on the analytical model different operating strategies
that either eliminate the need for real power exchange from
the power converter, minimize the amount of real power
exchange, or minimize the apparent power exchange have been
identified, depending on the sag level and load power factor. A
detailed computer simulation model for the system including
the dynamic and control elements has been developed and used
to verify the analytical model. A laboratory scale experimental
prototype system has also been used to confirm the results
from the computer simulations.
[1] ITI (CBEMA) Curve Application Note, Information Technology Industry
Council (ITI) Std., 2000.
[2] IEEE Recommended Practice for Evaluating Electric Power System
Compatibility with Electronic Process Equipment, IEEE Std. 1346-1998,
1998.
[3] D. O. Koval and M. B. Hughes, “Dynamic sag correctors: cost-effective
industrial power line conditioning,” IEEE Trans. Ind. Applicat., vol. 37,
no. 1, pp. 212–217, 2001.
[4] N. G. Hingorani, “Introducing custom power,” IEEE Spectr., vol. 32,
no. 6, pp. 41–48, 1995.
[5] N. H. Woodley, “Field experience with dynamic voltage restorer system,”
in IEEE Power Engineering Society Winter Meeting, Jan. 2000, pp.
2864–2871.
[6] K. N. Choma and M. Etezadi-Amoli, “The application of a DSTATCOM
to an industrial facility,” in IEEE Power Engineering Society Winter
Meeting, Jan. 2002, pp. 725–728.
[7] J. R. Clouston and J. H. Gurney, “Field demonstration of a distribution
static compensator used to mitigate voltage flicker,” in IEEE Power
Engineering Society Winter Meeting, Jan. 1999, pp. 1138–1141.
[8] W. E. Brumsickle, R. S. Schneider, G. A. Luckjiff, M. F. McGranaghan,
and D. M. Divan, “Canadian national power quality survey: frequency
of industrial and commercial voltage sags,” IEEE Trans. Ind. Applicat.,
vol. 337, no. 3, pp. 622–627, 1997.
[9] L. Gyugyi, C. D. Schauder, C. W. Edwards, and M. Sarkozi, “Apparatus and method for dynamic voltage restoration of utility distribution
networks,” U.S. Patent 5 329 222, July 12, 1994.
[10] Using Matlab, The MathWorks Inc, Natick, MA, 2003.
[11] Using Simulink, The MathWorks Inc, Natick, MA, 2003.
[12] G. Venkataramanan and B. Wang, “Dynamic modeling and control of
three phase pulse width modulated power converters in phasor domain,”
PESC 2004, digest submitted.
1451