A Modified Static Voltage Control Algorithm Based On Optimal Multiplier Method and Dynamic Generation Characteristics

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

A Modified Static Voltage Control Algorithm based on Optimal

Multiplier Method and Dynamic Generation Characteristics


Bo Gong Member, IEEE

Abstract—With the fast development of wide area measure- their maturity, fast computation time and practical for online
ments, communication techniques and dynamic security assess- applications.
ment (DSA) tools in power systems, dynamics based stability Nowadays, with the fast development of wide area mea-
control strategies are attracting more and more research atten-
tions as an effect way to handle complicated control objectives surements, communication techniques and dynamic security
and improve control accuracy. However, due to its heavily assessment (DSA) in power systems, dynamics based stabil-
relying on the time domain simulation, the computation burden ity control strategies are attracting more and more research
of these dynamic based methods has also been greatly increased. attentions. Dynamic controls are believed to be more accurate
On the other hand, static methods, which have been developed by guaranteeing the post-control stability in a longer term.
for years based on simplifications of the dynamic systems,
are constantly appreciated by industries for their maturity, Several approaches that based on time domain simulation and
fast computation time and practical for online applications. advanced control strategies have been proposed for long term
In this paper, in order to improve the accuracy of static voltage stability [8], [9]. However, due to its heavily relying
voltage stability control algorithms, the rules of simplifying on the time domain simulation, these dynamic based methods
a generator reactive power limitation, which is one of the induce greatly computation burden. This cause problems for
most important contributing factors for voltage collapses, has
been reviewed. A refined static model is proposed based on online applications as it may have difficulties to provide
considering this limitation characteristics in a dynamic control control solution in a real time.
framework. The static voltage stability control is computed by In this paper, in order to improve the accuracy of static
iteratively applying an optimal multiplier method. It shows that voltage stability control algorithms, the rules of simplifying
this method only add a small amount of computation burden a generator reactive power limitation has been reviewed. A
compared to the original static method, while it can improve the
control performance significantly during a dynamic process. An refined static model is proposed based on considering this
example voltage collapse is discussed to illustrate the algorithm limitation characteristics in a dynamic control framework.
and provide comparisons. The static voltage stability control is computed by iteratively
applying an optimal multiplier method. It shows that this
I. I NTRODUCTION method only add a small amount of computation burden com-
Voltage collapse is one of the major security concerns pared to the original static method, while it can improve the
for power system operations, especially when it triggers control performance significantly during a dynamic process.
cascading failures that lead to large scale system blackouts. A benchmark voltage collapse case (10 bus system) is used to
In reality, voltage collapse may occur within a wide range of illustrate the performance of this control strategy and provide
time-frames after the initial disturbances, from a few seconds some comparisons.
to a couple of hours. The longer term voltage instability The paper is organized as follows. Section II reviews static
involves some slow power transferring changes caused by voltage stability theory. Section III discusses static and dy-
load restoration and generator reactive power limitation. The namic generator reactive power limitation issues. Section IV
load that will gradually buildup could be the consequence focuses on static voltage stability control algorithm design
of load tap changer (LTC) or thermostat adjustments, and and improvement based on an optimal multiplier method. A
generator reactive power limitation may further exaggerate voltage collapse example using this control is provided in
this power transfer change and lead to voltage collapse. Section V and conclusion are presented in Section VI.
The relatively slow pace of this process provides sufficient II. S TATIC VOLTAGE S TABILITY T HEORY
time for implementation of operational decisions aimed at
preventing the collapse. Theoretical background for static voltage stability control
can be traced back to the earliest criterion for detecting
Static methods for voltage stability controls has been
voltage collapse proposed by Venikov [10]. Later, the appear-
studied for decades. Saddle node bifurcation theory, where
ance of theoretical explanation for voltage collapse by saddle
two equilibria coalesce and disappear [1], provided one
node bifurcation theory, where two equilibria coalesce and
widely accepted theoretical explanation for static voltage
disappear [1], provided the most widely accepted criterion
collapse. Many control designs based on this theory has
for designing a static minimal load shedding voltage control
been developed [2], [3], [4], [5], [6], [7]. These methods
strategy. In general, these static approaches are based on
have many advantages for industry applications such as
the following assumption: if a power system is modelled
Bo Gong is with Siemens PTI, Schenectady, NY, USA, 12305 (email: as a parameter dependent system, voltage collapse could
978-1-4577-1002-5/11/$26.00 ©2011 IEEE
bogong@siemens.com) happen when changes to the parameters drive system states
to the critical point lying on the saddle node bifurcation
hypersurface.
To describe the voltage collapse dynamics, a parameter
dependent power system modelled by differential algebraic
equation (DAE) is given as:
ẋ = f (x, y, λ)
0 = g(x, y, λ)
where x, y and λ correspond to dynamic states, algebraic
states and parameters respectively. The equilibrium of the
above equations are defined as (x ∗ (λ), y ∗ (λ)). If this equi- Fig. 1. Generator capability curve
librium is a stable equilibrium, any small perturbations to x
will not destabilize the system.
Synchronous generators in power systems are often pro-
Starting from an initially stable equilibrium (x 0 , y0 ). Let tected or controlled by various devices such as AVR (Au-
the parameter λ keep changing until the critical point of tomatic Voltage Regulator), GOV (Governor), OXL (Over
voltage collapse is reached. At this equilibrium, the nonlinear eXcitation limiter) and UXL (Under eXcitation limiter). In
system dynamics can be approximated by a linearized version power system operations, their cumulative effects based on
of the system equations as follow: the generator output P, Q are shown in figure 1, which
    
Δẋ fx fy Δx are often called generator capability curves. Curve 1 and 2
=
0 gx gy Δy indicate the field (rotor) and armature (stator) current limits.
Assuming gy is nonsingular we can eliminate Δy and obtain: Static capability curves are generator properties under
the assumptions of nominal voltages. In traditional power
Δẋ = (fx − fy gy−1 gx )Δx system operations, a generator’s reactive power capability
is often approximated by a constant value. This assumes
where fx , fy , gx , gy depend on λ.
curve 1 is flat (vertical in figure 1). This simplification,
Let: though bringing a great amount of convenience to system
 
fx fy studies, may introduce errors to represent this limitation
J = characteristics, especially during a voltage collapse process.
gx gy
in a dynamic process. Violations of the assumptions can be
explained by the mathematical definition of curve 1:
and
H = fx − fy gy−1 gx . vt2 2 Ef d vt 2
(1) P 2 + (Q + ) = ( ) . (2)
xq xd
By linear system theory, stability of the above linearized
system is determined by eigenvalues of the reduced Jacobian where vt is the terminal voltage.
matrix H. When all eigenvalues of H are negative or have Under normal conditions, generators’ terminal voltages are
negative real parts, the system is stable. By changing λ, locally controlled by AVRs. Curve 1 defined by equation (2)
the critical point will be the equilibrium where any of the with a constant field voltage E f d and a constant terminal
eigenvalues of Jacobian matrix H pass through 0 or the voltage vt ≈ 1 becomes a portion of a circle centered at
v2
imaginary axis. This was further simplified as the singularity (0, − xtq ). Along a voltage collapse process, however, the
of power flow Jacobian matrix, which can be much easier to generator terminal voltage can be controlled by OXL instead
justified [11]. This simplification serves as a basis for many of AVR. At this time, the terminal voltage of a generator will
later static voltage instability control approaches. no long be held constant. Curve 1 will become to depend on
values of vt .
III. G ENERATOR REACTIVE POWER LIMITS
A. Static generator capability curve B. Dynamic Over excitation limiter characteristics
Generators are the primary source of real and reactive
power in a power system. Reactive power is crucial to power
systems by providing voltage support. Typical synchronous
generators can produce reactive power at their terminals
from 31% of rated MVA (0.95 power factor generators) to
60% of rated MVA (0.8 power factor generators). In normal
conditions, a generator’s reactive power output is continu-
ously adjustable by an Automatic Voltage Regulator (AVR).
Along voltage instability process, however, the generator
reactive power capability may be limited by other factors. Fig. 2. Over excitation limiter model
16
An IEEE AC4A type OXL model is used in this paper Generator Capability Curve
Output PQ point
to illustrate the dynamic performance of OXL during the 14
voltage collapse process. Figure 2 shows the block diagram
of this model. By changing the field voltage set points, the 12
5 4
main objective of an OXL control loop is to regulate field 3
10 1 2
current to avoid overheated field windings. When a generator
10.87 5
is online, its OXL operates in either of the following 3 status:

P
4
8 10.88 3
• If the field current stays below a long term thermal 10.89

limit Ifmax1
d , output of OXL is 0. It adds no effects on 6
P10.9
2
generator outputs. 10.91
4
• If the field current exceeds the long term thermal limit 10.92

Ifmax1
d , but below a short-term thermal limits I fmax2
d , 2
10.93

10.94 1
OXL does nothing but waits until the heating effects
accumulate to a certain amount. Then it will send signal 0
6 8
6 8
10
10 12
Q 12
14
14
16
16 18
to AVR to bring field current down to the safe range. Q

• If the field winding current exceeds the short-term


Fig. 4. Generator PQ plot under OXL regulation
thermal limit Ifmax2
d , it immediately sends signal to
bring down the field current.
tions for voltage collapse controls. Field current measure-
ments, on the other hand, provide a better way to address
generator reactive power limitation in the dynamic control
IMax2
fd
framework.

Max1
IV. VOLTAGE S TABILITY C ONTROL A LGORITHM
I
fd

'T A. Optimal Multiplier Method and Static Voltage Control


As mentioned in section II, voltage collapse can be related
to a saddle node bifurcation where two equilibria coalesce
and disappear. Due to the difficulty of analyzing the reduced
Jacobian H in equation (1), the singularity of the power
flow Jacobian matrix is often used as an alternative indicator
for the voltage instability for a parameter dependent power
system. This implies the insolvability of the DAE equation
Fig. 3. Field current controlled by OXL set:

0 = f (x, y, λ)
Figure 3 shows field current under OXL regulation during
a voltage collapse process (same as the example discussed 0 = g(x, y, λ)
in section V). It can be observed that the field current was To avoid the divergence problem caused by traditional
regulated back to the long-term thermal limit I fmax1
d along Newton methods solving the above unsolvable DAE equa-
the process. Also, due to the long time constant of thermal tions, Tamura [13] proposed an optimal multiplier method
effects, generators may tolerate the over heating of field utilizing a variable Newton step. The optimal multiplier
windings for a short period ΔT . method will guarantee that each Newton step will conver-
For comparison purpose, the P-Q plot (damping out the gence and, especially when a feasible solution does not exist,
fast transients) of the same generator in the same process the algorithm will converge to a bifurcation point on the
is shown in figure 4. The generator was originally operated solution space surface, where the power flow Jacobian is
at point 1. Following a disturbance, the operating point is singular.
suddenly changed to point 2 and the by load restoration, Based on the optimal multiplier method, Overbye [5]
the reactive power output of generator keeps increasing until proposed a practical method to calculate the optimal voltage
point 3 was reached. At this time, the OXL is activated and stability control. In [14], it was proved that, at the solution
the field current was regulated to the long-term limits I fmax1
d . space boundary surface point x ∗ , the left eigenvector w ∗
Because of the terminal voltage reduction, the reactive power associated with the zero eigenvalue of Jacobian J(x ∗ ) is
Q will not stay at the limits but keeps dropping from point 4 normal to the surface. If the surface is flat, then the closest
to 5. Similar discussions of this kind of dynamic reactive point S 0 on the surface to the unsolvable point S can be
power limitation behavior can be found in [12]. calculated by:
The above dynamic process clearly shows that constant
reactive power limits of generators are problematic assump- S0 = S + [(fg (x∗ ) − S) ∗ w∗ ] w∗
Bus 1 Bus 4 1600MVA
Bus 7 T6 Bus 3 1094MW
where fg are the power flow equations. For a non-flat
hyper-surface, by iteratively applying the above equation, the Generator 1 T4 Generator 3
5000 MVA
solution may converge to such a closest point. 3632 MW
T1
Bus 8
Industrial load
After the closest bifurcation point is found, the sensitivity 3000 MW,
1800 MVAr

of the distance (between the unsolvable point and the bifur- 1500 MVAr

Bus 2
cation point) to some controls u is given [3] as: T2 LTC3 Residential and
commercial load

0
0 ∂(fg (x ) − S) 1
3000MW
Generator 2 T5

Φu = −w
Bus 10 Bus 6 Bus 9
2200 MVA
1500 MW Bus 5 300 MVAr
∂u 868 MVAr

This sensitivity is then used to compute the optimal voltage


Fig. 5. 10 bus voltage collapse test system.
stability controls.
B. Modified Algorithm Considering Dynamic Generation
Characteristics From the above equations, it can be obtained that
The transient of transmission network is quite fast so that it 
 dEq
is neglected in both static and dynamic methods. Therefore, Td0 = −(xd − xl )If d + Ef d
the static simplification rules of generators and loads play dt
very important roles in finding stabilizing controls. As many In a long term voltage instability process, neglecting the fast

other static voltage stability control methods, the static volt- transient of Eq , the relations between field voltage and field
age stability control method given in section IV-A assumed current can be established as:
PV bus (constant generated real power and terminal voltage)
for generators. In section III-B it has been discussed that a (xd − xl )If d = Ef d (3)
constant voltage assumption may be violated during a voltage
collapse process. This is one of the major contributing factors Even though OXL controls along a voltage collapse pro-
for the control error produced by static methods. A constant cess may render the P-V assumption of generators invalid,
field current provides a more accurate indication for reactive a more appropriate assumption for the static study of power
power limitation. systems can be made instead, utilizing the observation de-
In reality, generator’s field voltage as well as field current scribed above. This assumption is listed as follows:
may vary frequently to adapt to power flow variations in • If a generator’s field current magnitude is below its long
a system. At the same time, such saturation-like limitation term thermal limits, this generator is under AVR control.
caused by OXL defines an upper limit for generator field Its terminal bus can be approximated by a P-V bus.
winding current variation. This saturation effect is a relative • If a generator’s field current magnitude is staying at its
long term effect, considering generators can stand some short long term thermal limits, this generator is under OXL
period violations of the overheating limits caused by transient control. Its internal bus can be approximated by a P-V
phenomena. Such a field current limitation, shown in figure 3, bus, with constant field voltage magnitude determined
exists for most generator settings, since it is directly related by equation (3) and constant generator output real
to the overheating effects of field windings. Moreover, during power.
a voltage collapse process, assuming the field current keeps
In practice, a constant field voltage assumption means
increasing and is finally limited to I fmax1
d .
that the transmission network connecting generators and
To improve the control accuracy, the constant voltage
loads will be expanded by including the source impedance
assumption should be replaced by the constant field current,
of generators and use the internal bus voltage as the PV
as section III-B suggested. However, due to the fact that field
bus (neglecting the stator losses). With this expansion of
current are normally not metered in real time, a saturated
transmission network, same idea discussed in section IV-A
field voltage can be used instead. This can be justified by
can be used to compute static voltage stability control,
considering the generator dynamics given in [15] as follows:
bringing no significant changes to the algorithm itself but
  
incorporating the dynamic generation characteristics.
 dEq   xd − xd
Td0 = −Eq − (xd − xd )[Id − 
dt (xd − xl )2 V. E XAMPLE
 
(ψ1d (xd − xl )Id − Eq )] + Ef d In this section, a comparison between the original static

−1  xd − xd  voltage stability control and the modified method is discussed
If d = [E +  (−xl Id − ψd + Eq )]
xd − xl q xd − xl based on a benchmark 10 bus system.


xd − xl  x − xd
  The simple 10 bus system shown in Figure 5 is well
ψd = −xd Id +  Eq + d ψ1d established as a benchmark for exploring voltage stability
xd − xl xd − xl
issues [16], [4], [9]. The system has 3 generators and 2 loads.
1 Note that w0 , x0 are used for the closest boundary point and w∗ , x∗ LTC3 is a load tap changer (LTC) that automatically adjusts
are used for other points during each iteration step. its tap ratio according to the voltage magnitude at load bus 9.
Bus voltage for 10 buses system
1.05
Bus 3 1.2
Bus 9 Bus 3
1 Bus 6 Bus 9
Bus 6
1.15
0.95

0.9 1.1
Voltage (pu)

0.85

Voltage (p.u.)
1.05

0.8

1
0.75

0.95
0.7

0.65 0.9

0.6
0 20 40 60 80 100 120 140 160 180 200
time (second) 0.85
0 10 20 30 40 50 60 70 80 90 100
time (second)

Fig. 6. Bus voltage behaviour with no MPC control action.


Fig. 7. Controlled voltages for PV assumption

1.1
Bus 3
A. Voltage collapse Bus 9
Bus 6

An outage of any one of the feeders between buses 5 and 7 1.05

will lead to voltage collapse. This is illustrated in Figure 6 for


a line outage at 10 seconds. Voltages at buses 3, 6 and 9 are
shown. Bus 9 supplies the residential load, and is regulated 1
Voltage (p.u.)

by the LTC. Following the initial drop caused by the fault,


the voltage starts to recover as a consequence of LTC tap
0.95
ratio adjustments. When viewed from bus 6 though, this LTC
adjustment results in a load recovery process, which leads to
the voltage falling at that bus. All the other buses follow the 0.9

same trend of decreasing voltage, except generator terminal


buses, such as bus 3, which are regulated by automatic
voltage regulators (AVRs). At 35 seconds, increasing field 0.85
0 10 20 30 40 50 60 70 80 90 100
time (seconds)
current causes the reactive power limit at generator 3 to be
reached. The over-excitation limiter (OXL) at generator 3 Fig. 8. Controlled voltages based on modified method
reduces the internal field voltage. A sudden drop in the bus
voltages can be observed. After that, further LTC tap ratio
adjustments cause voltages across the system to drop. The VI. C ONCLUSION
field current and generator’s P-Q output are shown in figure 3
and 4 in section III. The long term field current limits are 26.3 In this paper, a refined static model is proposed based
pu. on considering dynamic generator reactive power limitation
characteristics. The static voltage stability control is com-
puted by iteratively applying an optimal multiplier method.
B. Static control performance It shows that this method only add a small amount of
computation burden compared to the original static method,
For the 10 bus system, with the same settings, a static while it can improve the control performance significantly
control computed based on PV assumption is 17.33% at during a dynamic process.
load 8. The post-control voltage trajectories are shown in
figure 7. The modified algorithm need a load shedding of R EFERENCES
12.63% at load 8, the system bus voltage is shown in figure 8. [1] I. Dobson and H.-D. Chiang, “Towards a theory of voltage collapse
in electric power systems,” Syst. Control Lett., vol. 13, pp. 253–262,
In order to boost generator terminal bus voltage to the 1989.
[2] T. Van Cutsem, “A method to compute reactive power margins with
nominal value, the original method sheds load aggressively, respect to voltage collapse,” Power Systems, IEEE Transactions on,
lead to the load voltage going above the upper limit of vol. 6, no. 1, pp. 145–156, 1991.
the LTC deadband. Such an unnecessary load shedding can [3] I. Dobson and L. Lu, “New methods for computing a closest saddle
node bifurcation and worst case load power margin for voltage
be effectively eliminated when considering the field voltage collapse,” Power Systems, IEEE Transactions on, vol. 8, no. 3, pp.
instead. The system are restored stable after the control. 905–913, 1993.
[4] P. Kundur, Power System Stability and Control. EPRI Power System
Engineering Series, McGraw Hill, 1994.
[5] T. Overbye, “Computation of a practical method to restore power flow
solvability,” Power Systems, IEEE Transactions on, vol. 10, no. 1, pp.
280–287, 1995.
[6] T. Van Cutsem, “An approach to corrective control of voltage insta-
bility using simulation and sensitivity,” Power Systems, IEEE Trans-
actions on, vol. 10, no. 2, pp. 616–622, 1995.
[7] Z. Feng, V. Ajjarapu, and D. Maratukulam, “A practical minimum load
shedding strategy to mitigate voltage collapse,” Power Systems, IEEE
Transactions on, vol. 13, no. 4, pp. 1285–1290, 1998.
[8] M. Zima and G. Andersson, “Stability assessment and emergency con-
trol method using trajectory sensitivities,” in Power Tech Conference
Proceedings, 2003 IEEE Bologna, 2003.
[9] I. Hiskens and B. Gong, “Voltage stability enhancement via model
predictive control of load,” in Proceedings of the Symposium on Bulk
Power System Dynamics and Control VI, August 2004.
[10] V. Venikov, V. Stroev, V. Idelchick, and V. Tarasov, “Estimation of
electric power system steady-state stability in load flow calculation,”
IEEE Trans. on Power Apparatus and Systems, vol. PAS-94, pp. 1034
– 1041, 1975.
[11] P. Sauer and M. Pai, “Power system steady-state stability and the load-
flow jacobian,” Power Systems, IEEE Transactions on, vol. 5, no. 4,
pp. 1374–1383, 1990.
[12] B. Gong, A. Pinheiro, and I. Hiskens, “Optimal load shedding with
the support of dynamic generation control in electric power systems,”
Proceeding of IEEE PES General Meeting, July, 2010.
[13] T. Tamura, K. Iba, and S. Iwamoto, “A method for finding multiple
load-flow solutions for general power systems,” IEEE PES Winter
Meetings, vol. A80, pp. 043–0, Feb. 1980.
[14] I. Dobson, “Observations on the geometry of saddle node bifurcation
and voltage collapse in electrical power systems,” Circuits and Systems
I: Fundamental Theory and Applications, IEEE Transactions on [see
also Circuits and Systems I: Regular Papers, IEEE Transactions on],
vol. 39, no. 3, pp. 240–243, 1992.
[15] P. W. Sauer and M. A. Pai, Power System Dynamics and Stability.
Prentice Hall, 1997.
[16] C. Taylor, “Concepts of undervoltage load shedding for voltage stabil-
ity,” Power Delivery, IEEE Transactions on, vol. 7, no. 2, pp. 480–488,
1992.

You might also like