Voltage Stability Preserving Invariants for Smart
Grids
Tamal Paul1, Harsha Ravindra2, Michael Steurer2, and Jonathan W. Kimball1
1 2
Department of Electrical and Computer Engineering
Missouri University of Science and Technology
Rolla, Missouri USA
tpzmb@mst.edu, kimballjw@mst.edu
Abstract— Voltage stability analysis is essential in any power
system. This paper addresses the voltage stability in a typical
smart grid type system with multiple independent entities. A
typical smart grid operation involves various loading excursions
(changes in power, both generated and consumed) undertaken
by all these independent entities. For a smooth functioning of
any generic smart grid type system, correct behavior of all these
independent entities must be preserved when one or more of
these entities are subjected to various loading levels. Correct
behavior of all the entities (sub-systems) will ensure correct
behavior of the overall system (smart grid). Invariants, if forced
to be true, ensure correct behavior on a subsystem level and
thus preserve the overall system correctness. An invariant is a
logical predicate on a system state that should not change its
truth value if satisfied by system execution [1]. This paper
derives an invariant that preserves voltage stability. This
invariant is based on an online indicator which is derived from
fundamental Kirchhoff’s laws and will predict the proximity of
voltage collapse at one or more entities in a smart grid. The
efficiency of the invariant in predicting voltage collapse has
been verified with simulations performed on a typical seven
node smart grid system. Thus an online monitoring of the
system parameters gives an indication of the system voltage
stability. The voltage stability invariant works for both static
and dynamic states. This method is also a fast and powerful tool
to predict the voltage stability margin of a generic smart grid
system by a simple monitoring of the system parameters.
I. INTRODUCTION
Voltage stability is of utmost importance to power system
operation. Drop in voltage levels poses serious concerns in
electric utilities and appliances. According to ANSI
standards, system voltages should not drop below 5% of their
nominal value for a satisfactory operation of the power
system [2-4]. Any study on voltage stability of a power
system should investigate how close the system voltage is to
collapsing, at what precise moment does the collapse occur,
what are the vulnerable areas in the system and what are the
factors contributing to this collapse.
A. Voltage Stability Studies
The concept of voltage stability is critical for the proper
functioning of a smart grid type system. With the ever
978-1-4799-7949-3/15/$31.00 ©2015 IEEE
Center for Advanced Power Systems
Florida State University
Tallahassee, Florida USA
hr09g@my.fsu.edu, steurer@caps.fsu.edu
growing increase in power demand, voltage stability
problems are also on the rise and pose a serious threat to
system operators. The problem of voltage stability can be
attributed to the inability of the system to cater to the reactive
power needs of its elements. It can this be described as a
reactive power problem [5].
Studies undertaken to analyze voltage stability require
careful investigation of a large number of system states and
contingency situations. This makes a steady state approach
seem more feasible. Literature studies reveal special
algorithms that target the voltage stability problem using
static methods [6]. However, such studies are labor intensive
and fail to provide sensitivity information. Certain operators
make use of the Q-V curve at some of the independent
entities to determine the proximity of voltage instability [7].
The major drawback of such an approach is that the excessive
focus on a small number of local components fails to provide
information about the overall system-level voltage stability
B. Voltage Instability
Voltage instability leads to a progressive (or periodic) fall (or
rise) in voltage of the system. As mentioned earlier, this loss
of voltage stability stems from the inability of the system to
supply the reactive power needed by one or more independent
entities. A voltage decay/fall in the system is usually caused
by an increasing in loading or change in system conditions at
one or more entities thereby causing an excessive increase in
reactive power demand which the system is unable to meet.
Thus the loading pattern of the individual components affects
the system-level voltage stability. This emphasizes the
importance of enforcing voltage stability constraints at the
sub-system level in order to ensure a proper functioning of
the overall system and eliminate the possibility of voltage
collapse in the system.
II. VOLTAGE COLLAPSE
One of the key indicators of a healthy system is steady
voltage levels at all of its sub-systems. The system is planned
to operate in such a way that whenever there is an increased
1
demand in load at one or more entities, the generators in the
system cater to that increased load by regulating the reactive
resources in the system. This remedial control action which
restores system voltage levels in the event of change in
loading conditions may include automatic response of the
generator excitation systems, automatic or manual excitation
of transformer taps and the switching of STATCOMs in the
circuit.
However under extreme conditions of loading, the
remedial action which earlier restored the voltage stability
might now very well reduce an already low voltage to
abnormally low levels, or may be too slow. Such low voltage
levels are not acceptable and might result in termination in
the continuity of supply of electricity. This chain of events
leading to a voltage collapse takes place in a matter of
seconds and causes the entire system to shut down. This
poses a serious threat to power system engineers and poses a
challenge to them to identify conditions or vulnerable areas in
the system which might lead to a voltage collapse at one or
more entities and thus shut the entire system down.
Voltage collapse is caused by insufficient reactive power,
either a lack of enough reactive power in the system to
deliver the active power or too much consumption of reactive
power within the system itself. This happens in stressed
power systems. A system may enter the state of voltage
collapse when an increase in load demand at one or more
subsystems or change in system conditions causes an
unrecoverable decrease in one or both bus voltages. Usually it
is the variation in load that is the primary culprit behind
voltage collapse.
III. VOLTAGE COLLAPSE PROXIMITY IDENTIFICATION IN A
SMART GRID
Once the voltage at one or more entities (subsystems) starts to
fall, no remedial action can be taken to revert the system back
to its normal state of operation. In other words, once the
system has entered the state of voltage collapse, remedial
measures are either too slow or too inadequate to thwart the
collapse. Thus identifying the proximity of the system to the
voltage collapse point is of crucial importance.
The voltage collapse proximity indicator L [8] is a
quantitative measure to predict how far the current state of the
system is from the voltage collapse point. The value of L
ranges between 0 (no load) and 1 (voltage collapse). This
means as L approaches 1, the system approaches the brink of
voltage collapse. The voltage collapse proximity indicator for
any entity is mainly influenced by the load at that entity itself
and the influences from other entities. The voltage collapse
proximity indicator for any load bus ( L j ) for any load bus j
can be given as [9]:
S j'
Lj = (1)
V j2Y jj
2
N Z ji *Si
S j' = S j + ¦ Vj (2)
i =1 Z jj *Vi
i≠ j
where N is the total number of load buses, S j and Si are the
complex power at buses j and i, V j and Vi are the voltages
at buses j and i, Y jj is the jj th
element of the bus admittance
matrix of the system and Z *jj and Z *ji are the conjugate of
the jj th and ji th elements of the bus impedance matrix of the
system. The maximum value of the indicator for all the
independent entities is called the “System Indicator” and it
indicates the health of the overall system.
LSystem = Maxi∈{1"N} ( Li ) (3)
The indicator for voltage collapse proximity at a load bus is
mainly influenced by the load at bus j itself and the influences
from other load buses. The indicator is influenced by a
change in load at a load bus. Thus the voltage collapse issue
is not restricted only locally; it is a system level issue. The
maximum value of the indicator in the system thus signals the
voltage collapse of the system.
IV. INVARIANT FORMULATION TO PRESERVE VOLTAGE
COLLAPSE IN A SMART GRID
An invariant, as the term is used in the present work and in
[1], is a logical predicate that must remain true throughout
system execution. This use of the term derives from computer
science concepts and allows for automatic verification that
the system is operating correctly. The present goal, as in [1],
is to derive an invariant that could be used by a cyber
controller in the smart grid to ensure that the active and
reactive power flow settings of the system entities do not lead
to voltage collapse.
When the load is increased by a certain entity, the
indicator at the corresponding bus and/or at other buses starts
to increase. To avoid voltage collapse, the per-unit voltage at
any bus should always be greater than the indicator for any
bus. There comes a point where the indicator for one bus
intersects the per-unit voltage of a certain bus, either its own
or another bus in the system. This point is called the “online
monitored voltage collapse indication point” where the
indicator approaches unity thereby indicating that the voltage
collapse is imminent. Ideally, the “online monitored voltage
collapse indication point” should precede the voltage collapse
point, or even the point where the voltage at one or more
buses drops below 5% of the nominal level as per ANSI
standards.
Thus the invariant to preserve system level voltage
stability can be formulated in words as: The maximum
indicator for all the independent entities must be always less
than the minimum per-unit voltage for all the independent
entities in the smart grid. Or mathematically,
{IV : Lsystem < Vi ∀i} (4)
Thus it is possible to perform an online check as long as the
system is running to monitor the system health in terms of
voltage stability and be forewarned should the invariant
indicate an impending voltage collapse.
V. MODEL DESCRIPTION
The model was simulated using RSCAD as shown in Fig. 1.
The power system model used for the study is the seven
section FREEDM feeder system. The system derives its name
from seven primary line sections. The feeder has 7 lumped
loads connected through 7 load SSTs. The feeder is rated at
12.47 kV. The maximum load on the feeder is 2 MVA. The
feeder has some peculiar features pertaining to studies in
FREEDM. The system is a looped system. It is also an
islanded feeder. The loads are served primarily by two diesel
generators.
Although the feeder is looped, it could be operated in
radial mode. The two diesel generators are rated at 1.4 MVA
each. One of them operates in droop mode (G2) while the
other operates in isochronous mode (G1). As mentioned
before, the feeder primary is made up of seven line sections.
The impedance of each line section is identical. Each load
SST is rated at 285 kVA. On the secondary of each load SST,
there is a lumped load, lumped distributed generation (DG)
and storage connected. The model parameters are given as:
• Generators: Rated at 1.4 MVA, 380 V line-to
neutral.
• Transformer: Rated at 1.5 MVA, 0.38kV:12.47kV
• Lines: 0.77039 ȍ (Positive Sequence resistance),
1.967 mH (Positive Sequence inductance),
2.310927 ȍ (Zero Sequence resistance), 5.9 mH
(Zero Sequence inductance)
The droop generator is set to run at 400 kW with 5% droop.
VI. SIMULATIONS AND RESULTS
Fig. 1: Single line diagram of the 7-bus system
3
A. Case I: Increase the loading at one bus (j) and observe
the voltage collapse of that bus. Monitor the stability
indicator for that bus ( L j ) and another bus ( Li )
The system conditions are given as:
1) The generation at each bus is kept at 0; thus assuming
no local generation.
2) The loading at buses 2-7 are kept at 205 kW.
3) Loading at bus 1 is increased throughout the simulation
from 205 kW to 1605 kW.
4) The total load on the system is the summation of the
load at each bus.
5) Generator 1 tries to supply the excess load in the
system
The loading at bus 1 is increased; keeping the load at all the
other buses fixed at 205 kW. The indicator for bus 1 (L1) is
observed in Fig. 2(a). The indicator for bus 3 (L3) is observed
in Fig. 2(b). Thus j=1 and i=3.
Results are indicated in Fig. 2. The solid blue line
indicates bus 1 per unit voltage and solid red line indicates L1
in Fig 2(a). The solid blue line indicates per unit voltage of
bus 1 and the solid green line indicates L3 in Fig. 2(b).
(a)
(b)
Fig. 2. Change in voltages and indicators as loading as bus 1
increases, (a) V1 and L1 and (b) V1 and L3.
 Bus 1 collapses at around 34 seconds when the loading at
bus 1 is 1.404 MW. The per-unit voltage of bus 1 at that
instant is less than 0.95 p.u (does not meet ANSI mandated
5% voltage deviation standards). L1 however intersects the
per unit voltage at around the 24 second instant when the
loading is 1.005 MW. This is the “online monitored voltage
collapse indication point” where L1 approaches unity. If
loading is continued beyond this point, the voltage collapses
shortly and eventually the system totally crashes.
Although the indicator for bus 1 indicates voltage
instability when the loading at bus 1 reaches the voltage
collapse point, the other bus indicator (L3) gives a false sense
of voltage stability by not indicating a violation of the
condition for voltage stability. This confirms the importance
of choosing the maximum value of L for determining the
(a)
voltage collapse point of the overall system.
Thus the online monitored indicator can signal when the
system is on the brink of voltage collapse and can thus act as
a guard/invariant to prevent the system into entering such an
undesirable state.
B. Case 2: Increase the loading at one bus (i) and observe
the voltage collapse of another bus (j). Monitor the
stability indicator for the other bus ( L j ) and the bus
which is being loaded ( Li )
The system conditions are identical to that of Case 1. The
loading at bus 1 is increased; keeping the load at all the other
buses fixed at 205 kW. The indicator for bus 1 (L1) is
observed in Fig. 3(a). The indicator for bus 4 (L3) is observed
in Fig. 3(b). Thus j=4 and i=1. (b)
The solid blue line indicates bus 4 per unit voltage and Fig. 3. Change in voltages and indicators as loading as bus 1
solid red line indicates L4 in Fig. 3(a). The solid blue line increases, (a) V4 and L4 and (b) V4 and L1.
indicates per unit voltage of bus 4 and the solid green line
indicates L1 in Fig. 3(b).
It is seen that bus 4 collapses at 34.67 seconds when the
loading at bus 1 is 1.428 MW. The per unit voltage of bus 4 at
that instant is less than 0.95 p.u. L4 never intersects the per
unit voltage curve; L1 however does.
Although the indicator for bus 4 does not indicate voltage
collapse when its own bus reaches the voltage collapse point
due to loading at bus 1, the other bus indicator (L1) indicates
an invariant violation. This re-iterates the importance of
choosing the maximum value of L for all the buses to
determine the voltage collapse point of the overall system.
For both cases 1 and 2, the voltage starts to collapse
because of insufficient reactive power in the system. This is
shown by Fig. 4 which depicts the generator 1 and 2 reactive
power. It can be seen while generator 1 absorbs reactive
power for these two cases, generator 2 supplies reactive
power; thereby keeping the overall reactive power generation
positive. However as the loading is further continues to
increase, generator 2 is unable to provide the reactive power
to compensate for the increased consumption of reactive
power of the system. Hence the voltage starts to collapse at
this point at one or more buses when the supplied reactive
Fig. 4. Plot of generator 1 and 2 reactive power against time
with increased loading at bus 1

4
power is greater than the reactive power needed to maintain
one or more bus voltages.
When the system is stressed, the droop generator supplies
reactive power to maintain the bus voltages. Case 3 is a study
of the system when the droop generator has been removed. In
the absence of a droop generator, the system becomes more
vulnerable than before and cannot be subjected to its previous
loading levels. If the loading is increased at one of the buses
(the loading levels being much smaller than the previous two
cases given that the system is already vulnerable in the
absence of an extra reactive power supplying generator),
generator 1 tries to supply the reactive power need of the
system as long as it can; but ultimately fails thereby resulting
in a total system crash as shown in Case 3.
C. Case 3: Increase the loading at one bus (j) in the system
without any droop and observe the voltage collapse of
that bus. Monitor the stability indicator for that bus
( L j ) and the point where the voltage collapses.
The system conditions are given as:
1) The generation at each bus is kept at 0; thus assuming
no local generation.
2) The loading at buses 2-7 are kept at 0 kW.
3) Loading at bus 1 is increased throughout the simulation
from 205 kW to 1500 kW.
4) The total load on the system is the summation of the
load at each bus.
5) Generator 1 tries to supply the excess load in the
system.
In the absence of the droop, the system collapses as can be
seen for a much lesser load as demonstrated in Fig. 5. This
crash is because the second generator no longer supplies the
deficit reactive power for the system. The only generator in
the model supplying reactive power quickly runs short of
reactive power causing a total system crash as shown in Fig.
6. The indicator for bus 1 signals the imminent occurrence of
Fig. 5. Plot of V1 and L1 against time with increased loading at
bus 1 for the system with no droop generator.
5
Fig. 6. Plot of generator 1 reactive power against time with
increased loading at bus 1.
the crash as it intersects the voltage curve prior to the system
crash.
The invariant signals the nearing of the system to voltage
collapse well ahead of time. Thus in a transaction based [10]
power system, each and every customer’s (sub-system’s)
obligation can be determined using this online system state
monitoring indicator.
VII. CONCLUSIONS AND FUTURE WORK
An invariant which predicts the proximity of voltage collapse
in a smart grid type setting has been presented. This invariant
is based on a real time online monitoring of system
parameters. The invariant can be suitably modified to provide
insight about the system voltage stability in the absence of
current/live system values. The efficacy of the invariant in
predicting voltage collapse has been verified with
experimental simulations. Should loading be continued any
further, the system does go on to collapse. The invariant thus
is useful in locating the vulnerable locations in the system.
The voltage collapses because of a deficit of reactive
power supply in the system as the reactive power needed to
maintain the sub-system voltages within acceptable levels
increases with increased loading at the buses. Generator 2
provides a “stabilizing” factor to the model by being a
provider of reactive power. In the absence of the droop
generator, the vulnerability of the system increases, its
tolerance towards loading decreases and it crashes at a much
lower load than it would have had the droop generator been
available. Thus the droop generator is important in widening
the satisfactory range of operation of the system and catering
to higher loads.
The invariant currently relies on online measurements of
the system states to give an effective assessment of the
proximity of the system to its collapse. Future work (in
progress) will explore what assumptions/ modifications can
be imposed on the existing indicator for it to function when
one or more system state measurements might not be
available for a variety of reasons. Unavailability of system
states may be due to absence of current/updated values or
deliberate corrupt/bogus values. In the absence of reliable
data, the indicator must be modified accordingly so that it can
still predict voltage collapse just like it does now when real
time correct system state data is available.
ACKNOWLEDGMENTS
This work was supported by the National Science
Foundation under award EEC-0812121, the Future
Renewable Electric Energy Delivery and Management
Systems Center (FREEDM).
REFERENCES
[1] Paul, T.; Kimball, J.W.; Zawodniok, M.; Roth, T.P.; McMillin,
B., Chellappan, S., "Unified Invariants for Cyber-Physical
Switched System Stability," Smart Grid, IEEE Transactions on,
vol.5, no.1, pp.112,120, Jan. 2014.
[2] ANSI C84.1-2011.
[3] Kusko, A. and Thompson, M. (2007). Power Quality in
Electrical Systems. New York: McGraw-Hill.
[4] Pacific Gas and Electric Company. (1999). Voltage Tolerance
Boundary.
[5] J. A. Diaz de Leon II, C. W. Taylor, “Understanding and
Solving Short-Term Voltage Stability Problems,” IEEE Power
Engineering Society Summer Meeting, 2002, pp. 745-752.
[6] I Aumuller, C.A.; Saha, T.K., "Investigating the impact of
Powerformer on voltage stability by dynamic simulation,"
Power Systems, IEEE Transactions on, vol.18, no.3,
pp.1142,1148, Aug. 2003.
[7] Huang, G.M.; Kun Men, "Contribution allocation for voltage
stability in deregulated power systems," Power Engineering
Society Summer Meeting, 2002 IEEE , vol.3, no.,
pp.1290,1295 vol.3, 25-25 July 2002.
[8] Wood and B.F Wollenberg, “Power Generation, Operation and
Control”, John Wiley, New York, 1996.
[9] I. Kumaraswamy, W.V Jahnavi, T. Devaraju, P.Ramesh, “An
Optimal Power Flow (OPF) Method with Improved Voltage
Stability Analysis”, Proceedings of the World Congress on
Engineering, London, UK, July 4-6, 2012.
[10] Kazari, H.; Abbaspour-Tehrani Fard, A.; Dobakhshari, A.S.;
Ranjbar, A.M., "Voltage stability improvement through
centralized reactive power management on the Smart Grid,"
Innovative Smart Grid Technologies (ISGT), 2012 IEEE PES,
pp.1,7, 16-20 Jan. 2012.