Professional Documents
Culture Documents
Power System Damping Control Via Power Injections From Distributed Energy Storage
Power System Damping Control Via Power Injections From Distributed Energy Storage
Power System Damping Control Via Power Injections From Distributed Energy Storage
Abstract—Inter-area oscillations are present in all power feedback from wide-area measurements of the generators’ ma-
systems dispersed over large areas and can have detrimental chine speeds. Naturally, injections from ES devices in different
effects limiting transmission capacity or even causing blackouts. locations throughout the system result in different transient
The availability of wide-area measurements in power systems
has enabled damping of inter-area oscillations using distributed responses and, thus, different damping performance. With
control methods and system components, such as energy storage multiple ES devices available throughout the system, there
devices. We investigate the performance of damping control may be scenarios in which it is advantageous to use smaller
enabled by energy storage devices distributed throughout an current injections from more ES devices distributed throughout
example two-area power system assuming the availability of wide- the system rather than using larger current injections from
area measurements of generator machine speeds. The energy
storage devices are capable of injecting active power into the fewer ES devices. We quantify performance for these types of
system in order to damp inter-area oscillations that occur after scenarios and provide insights for how best to utilize ES de-
a fault in the system. An analysis of the linearized system vices for damping control in this example system. Specifically,
and several simulations of the nonlinear system with multiple we perform a linear analysis as well as nonlinear simulations
combinations of controlled power injections from energy storage of the two-area example power system with damping control
devices are performed. From the results, we quantify and discuss
how damping performance depends on the location, size, and enabled by active power injections from multiple ES devices;
number of injections. we quantify the damping performance of several combinations
of ES devices providing injections, discuss how performance
I. I NTRODUCTION depends on the number, sizes, and locations of injections, and
show that there are diminishing improvements on damping
Power systems consist of complex interconnections of non- performance as the size of injections increase.
linear components, possibly sparsely distributed across wide Several authors have previously investigated the use of ES
geographic regions. This means that power loads and gen- devices for damping control of power systems similar to the
eration centers may be geographically separated by large dis- one considered in this work. Damping inter-area oscillations
tances. In these systems, an event, such as a fault at one of the in power systems with Superconducting Magnetic Energy
buses, results in swings in the power transfer between regions. Storage (SMES) is studied in [12], [13]. Damping with an
These power swings are called inter-area oscillations [1], [2], ES device using particle swarm optimization and heuristic
and they occur in power systems around the world. Damping dynamic programming is discussed to be better than PSS
inter-area oscillations is crucial for maintaining a secure and or FACTS enabled damping in [14] for the same example
reliable power grid, and failing to do so can have severe system and signals that we consider. An ES system based on
consequences, such as the blackouts experienced throughout UltraCapacitor technology is utilized for damping control via
western North America in 1996 [3]. real power modulation in [11]. The same authors show that
Several methods for implementing damping control that the optimal placement for an ES device to provide damping
mitigate the effects of inter-area oscillations have been pro- in a power system is in the area with lower inertia [15]. The
posed. Power System Stabilizers (PSSs), utilizing local mea- authors of [16] consider a similar system with only one ES
surements, have historically been used to implement damping device in each area and show that the inter-area oscillations
control for this task. More recently, the use of remote signals can be effectively damped and that the best locations for the
with PSSs has been shown to be advantageous [4]. The ES devices along the tie line are closer to the generators. In
availability of system-wide information via remote signals has this work we extend these previous results by investigating
enabled the use of other system components for damping injections from multiple ES devices distributed throughout the
control, such as Flexible Alternating Current Transmission system. More broadly, other applications for battery ES in
System (FACTS) devices, High Voltage DC (HVDC) lines, power systems are discussed in [17]. Several power and energy
and Energy Storage (ES) [5]–[11]. applications of ES as well as optimal energy management with
In this work, we specifically investigate damping control ES are discussed in [18], and multiple applications and value
enabled by active power injections from ES devices distributed opportunities for ES are highlighted in [19].
throughout the system. The damping control law is based on This paper is organized as follows. First, we formulate the
problem by introducing the example two-area power system
D. Copp, F. Wilches-Bernal, and D. Schoenwald are with Sandia National
Laboratories, Albuquerque, NM USA (e-mail: dcopp@sandia.gov). I. Gyuk and damping control law under consideration in Section II.
is with the U.S. Department of Energy, Washington, DC USA. Next, we present a linear analysis of the resulting closed-
loop system in Section III. Numerical results and damping TABLE I
performance for the nonlinear system are given in Section IV. M ODEL PARAMETERS
Finally, Section V contains concluding remarks and directions Generators:
for future work. Parameter Description Value Units
Pr power rating 900 MVA
Xd direct axis synchronous 1.8 per unit
II. P ROBLEM F ORMULATION reactance
Xq quadrature axis 1.7 per unit
synchronous reactance
A. Two-Area Power System Model Xl leakage reactance 0.2 per unit
Xd0 transient reactance 0.3 per unit
We analyze the two-area power system shown in Figure Xq0 transient reactance 0.55 per unit
Xd00 subtransient reactance 0.25 per unit
1 that has two synchronous generators and a load in each Xq00 subtransient reactance 0.25 per unit
area. This model, originally proposed in [20], is augmented to Ra armature resistance 0.0025 per unit
0
include six ES devices distributed throughout the system. The Td0 open-circuit time constant 8.0 seconds
0
Tq0 open-circuit time constant 0.4 seconds
synchronous generators are located at buses 1, 2, 3, and 4. The 00
Td0 open-circuit subtransient 0.03 seconds
loads are located at buses 18 and 19, and the ES devices are time constant
located at buses 12, 13, 14, 15, 16, and 17. The ES devices 00
Tq0 open-circuit subtransient 0.05 seconds
are capable of injecting and absorbing active power into the time constant
H1 , H2 inertia constant 6.5 per unit
system in order to damp inter-area oscillations. To excite the H3 , H4 inertia constant 6.175 per unit
inter-area oscillation of the system, we apply a three-phase to KD damping coefficient 0 per unit
ground fault and investigate the damping performance when Transmission:
Parameter Description Value Units
different combinations of ES devices provide active power
r line resistance 0.0001 per unit/km
injections. For example, we quantify the differences in control xL line reactance 0.001 per unit/km
effort (magnitude of power injected) and transient system bC line charging 0.00175 per unit/km
response when all six of the ES devices provide injections zT transformer impedance 0.0167 per unit/km
System:
versus when only ES devices 1 and 6 provide injections. Parameter Description Value Units
With each synchronous generator modeled using a subtran- Sb system base MVA 100 MVA
sient model, the system dynamics are given by Sf system frequency 60 Hz
0
sient terminal voltage Eq,i , direct axis subtransient flux link- and Ω2 (t) = ω3 (t)+ω
2
4 (t)
), respectively. These feedback signals
00 0
age ψd,i , direct axis transient terminal voltage Ed,i , and are used because the generators in a given area are often
00
quadrature axis subtransient flux linkage ψq,i . The vector coherent, so the difference between the average machine
of active power injections at time t is denoted by u(t) = speeds in the two areas is a good measurement of the inter-area
[u1 (t) u2 (t) u3 (t) u4 (t) u5 (t) u6 (t)]> ∈ R6 , and the output oscillation. Moreover, our objective is to damp the oscillations
of the system at time t is denoted by y(t) ∈ R6 . The in machine speeds in order to improve stability of the system
output for generator i is its angular velocity, so the (possibly while also reducing wear and tear on the mechanical compo-
remote) measurements of the generators’ angular velocities are nents. The number of injections in areas 1 and 2 are denoted by
available for feedback. The function g(·) captures the algebraic I1 and I2 , respectively, and kd is the damping control gain.
aspects of the model. For brevity we do not explicitly write Therefore, the magnitude of each injection is scaled by the
down the functions f (·) and g(·) in (1), but more details on number of injections in the area in which it is located. In this
these dynamics can be found in, e.g., [2], [21]. way, the sums of the magnitudes of the injections in each area
The model parameters are chosen to be the same as in are the same regardless of the number of injections in that area.
Example 12.6 of [2] and are given in Table I. Except for the This is done to determine how performance changes when
inertia constants H1 , H2 , H3 , and H4 , all of the parameter injecting power in one location versus injecting the same total
values are the same for all four of the generators. Saturation amount of power but through several injections distributed
effects are neglected. throughout the system.
Fig. 1. Example two-area power system.
kd kd kd kd
0 2I1 05 2I1 05 − 2I1
05 − 2I1
010
kd kd kd kd
0
2I1 05 2I1 05 − 2I1
05 − 2I1
010
kd kd kd kd
0
2I1 05 2I1 05 − 2I 05 − 2I 010
K = 1 1
,
kd kd kd kd
0 − 2I 2
05 − 2I 2
05 2I2 05 2I2 010
kd kd kd kd
Fig. 2. Block diagram of current injections from ES devices. 0 − 2I 2
05 − 2I 2
05 2I2 05 2I2 010
kd kd kd kd
0 − 2I 2
05 − 2I 2
05 2I2 05 2I2 010
Figure 2 shows a block diagram of the model for the active and
power injections from the ES devices. In Figure 2, Ωi denotes
e1 0 0 0 0 0
the average of the angular velocities of the generators in area 0 e2 0 0 0 0
i, Ωj denotes the average of the angular velocities of the 0 0 e3 0 0 0
generators in the other area, and T denotes the response time S=
0
.
0 0 e4 0 0
of the ES devices. The transfer function 1/(1 + T s), where 0 0 0 0 e5 0
s is the Laplace transform variable, acts as a first-order low- 0 0 0 0 0 e6
pass filter that captures the dynamics of the injection. In this
The notation 0n in the matrix K denotes a row vector with n
work, we choose T = 0.05 [s], which is representative of
elements all equal to zero. In the matrix S, each variable ei ,
the response time of some types of energy storage, such as
for i ∈ {1, 2, 3, 4, 5, 6}, is equal to one if the ith ES device
Superconducting Magnetic Energy Storage (SMES) [10]. The
is providing an injection or is equal to zero otherwise. Each
signal ūi (t) then denotes the actual power injection of ES
row of the matrix SK sums to zero, so the control inputs are
device i into the system.
zero when the difference between the average machine speeds
III. L INEAR A NALYSIS is equal to zero. This also means that the power injected and
absorbed by the ES devices is energy neutral throughout the
Because we are interested in small signal stability, we can time the control is active (i.e., assuming no losses, the state-
linearize the nonlinear dynamics (1) around the steady state of-charge of the ES does not change). Finally, the damping of
values and get a good approximation of how the system the closed-loop system (3) can be evaluated by computing the
behaves due to inter-area oscillations. The linearized dynamics eigenvalues of the matrix Ā.
can be written as
IV. N UMERICAL R ESULTS
ẋ(t) = Ax(t) + Bu(t), (2) In this section we present simulation results for the example
two-area power system shown in Figure 1. We analyze a 25
where x(t) and u(t) denote the state and input vectors,
second window in which the system is excited by a three-phase
respectively, as described after (1). The matrix A ∈ R30×30
to ground fault at bus 7 at t = 0.5 [s], and the fault is cleared
encodes the system dynamics, and B ∈ R30×6 is the input
at time 0.505 [s]. After the fault, we investigate the effect
matrix.
that active power injections from multiple combinations of ES
The feedback control can be written as a function of the
devices have on the system’s damping. There are 63 different
states so that the closed-loop dynamics can be written as
ES injection combinations available, but we only present 25
ẋ(t) = Āx(t), (3) combinations that are representative of the options available.
All simulations of the nonlinear system are performed using
where the Power System Toolbox [22] in MATLAB.
The initial conditions of the system after solving the load
Ā = A + BSK, flow are as follows: The initial loads in areas 1 and 2
are (PL1 , QL1 ) = (12.67, 1.00) [pu] and (PL2 , QL2 ) = TABLE II
(14.67, 1.00) [pu], respectively, and the generators are loaded P ERFORMANCE R ESULTS ( ALL VALUES ARE IN PER UNIT )
as follows (all values are in per unit): ES kΩ1 − Ω2 k2 kukmax
Case injections kd = 200 500 kd = 200 500
P1 = 7.00 Q1 = 2.09 V1 = 1.03∠0.87° 1 1,6 0.00299 0.00239 0.0549 0.1349
P2 = 7.00 Q2 = 2.93 V2 = 1.01∠ − 8.96° 2 1,2,5,6 0.00309 0.00247 0.0275 0.0676
3 1,3,4,6 0.00320 0.00258 0.0276 0.0678
P3 = 7.01 Q3 = 1.52 V3 = 1.03∠ − 6.8° 4 1,2,3,4,5,6 0.00320 0.00258 0.0184 0.0452
5 2,5 0.00320 0.00258 0.0552 0.1356
P4 = 7.00 Q4 = 1.55 V4 = 1.01∠ − 16.50° 6 3,4,6 0.00333 0.00271 0.0553 0.1362
7 3,4,5,6 0.00333 0.00271 0.0553 0.1362
Table II shows damping control performance results from 8 1,3,4 0.00334 0.00271 0.0553 0.1362
a numerical analysis of the two-area power system shown in 9 1,2,3,4 0.00334 0.00271 0.0553 0.1362
10 2,3,4,5 0.00334 0.00271 0.0277 0.0681
Figure 1 after a three-phase to ground fault occurs at bus 7 11 3,4,5 0.00342 0.00278 0.0554 0.1364
with active power injections from ES devices, as shown in 12 2,3,4 0.00342 0.00279 0.0554 0.1364
Figure 2. Results are shown for two values of the control gain, 13 6 0.00344 0.00282 0.0559 0.1367
14 1 0.00349 0.00286 0.0555 0.1369
kd = 200 and kd = 500. These values can be chosen based 15 3,4 0.00351 0.00287 0.0555 0.1367
on the size of the ES devices; we choose two values to show 16 5,6 0.00355 0.00291 0.0278 0.0685
how damping performance improves with larger injections. 17 1,2 0.00359 0.00296 0.0278 0.0686
18 4,6 0.00366 0.00302 0.0278 0.0687
The rows of Table II are organized by resulting performance 19 4,5,6 0.00366 0.00302 0.0186 0.0458
so that the combination of ES devices that produce the best 20 5 0.00367 0.00302 0.0557 0.1373
damping are at the top, and the result with no injections is at 21 1,3 0.00371 0.00307 0.0279 0.0688
22 1,2,3 0.00372 0.00307 0.0186 0.0459
the bottom. The order is not affected by considering different 23 2 0.00372 0.00307 0.0557 0.1375
values for kd . The first column gives a case number, and the 24 4,5 0.00381 0.00316 0.0279 0.0689
second column shows the combination of ES injections used 25 2,3 0.00387 0.00321 0.0279 0.0690
Base none 0.00549 0.00549 0 0
for that case. The third and fourth columns show the value of
the 2-norm of the difference between Ω1 and Ω2 , for kd = 200 8 8
and kd = 500, respectively, from time 0 to 25 seconds. This
metric quantifies damping performance (i.e., a smaller value of 6 6
the 2-norm means that the system is better damped). Finally,
the fifth and sixth columns show the value of the max-norm of 4 4
the input vector u for the entire simulation (i.e., the maximum
2 2
absolute value of u), thereby quantifying the size of the largest
injection commanded by the damping controller.
0 0
The A and B matrices in the linearized dynamics (2) can -4 -3 -2 -1 0 -4 -3 -2 -1 0
be found using the Power System Toolbox [22], and then the
(a) kd = 200 (b) kd = 500
eigenvalues of Ā can be determined. The eigenvalues for four
cases are shown in Figure 3. For those same four cases, Figure Fig. 3. Least damped eigenvalues of Ā when ES devices 1 and 6 are used
4 shows simulation results of the nonlinear system. Figure 5 (Case 1), when all six ES devices are used (Case 4), and when ES devices
shows the corresponding injections that produce the responses 3 and 4 are used (Case 15). The least damped eigenvalues corresponding to
the inter-area oscillation move farther to the left as kd increases. Eigenvalues
shown in Figure 4. Naturally, increasing the control gain kd symmetric about the Real Axis are not shown.
from 200 to 500 increases the size of the injections, and the
damping performance improves.
All of these results are consistent with the results in [15];
A. Discussion better performance is achieved when using injections from ES
Any injections from ES devices improve the damping per- devices in the area with lower inertia. In this example, area
formance regardless of how many or where they are located. 2 has lower inertia, so using only an injection from ES 6
In order to get the best improvement in damping performance, produces better results than using only an injection from ES
however, the number of injections and their locations are 1. Similarly, using only an injection from ES 5 produces better
crucial as, depending on the choice of these options, damping results than using only an injection from ES 2. The same is
performance may improve 29.5% to 45.5% when kd = 200 true for multiple injections in each area; injections from ES
and 41.5% to 56.5% when kd = 500. Moreover, carefully 5 and ES 6 provide better results than injections from ES 1
choosing the number and locations of smaller injections can and ES 2, etc. This is further confirmed by computing the
produce better damping performance than larger injections that eigenvalues of Ā for these cases. Figures 6 and 7 show the
are not carefully chosen. For example, injections from ES 1 corresponding eigenvalues and numerical results for some of
and 6 with kd = 200 produce better results than injections these cases, respectively.
from ES 4 and 6 with kd = 500, and the largest injections are As we saw previously, there is a natural correlation between
1.25 times smaller. the size of injections and the damping performance. In general,
-4
10
3 0.06
2 0.04
1 0.02
0 0
-1 -0.02
-2 -0.04
-3 -0.06
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
(b) kd = 500
(b) kd = 500
Fig. 5. Power injections when ES devices 1 and 6 are used (Case 1), when
Fig. 4. Difference between the average machine speed in area 1 and area all six ES devices are used (Case 4), and when ES devices 3 and 4 are used
2 after the fault. Injections from only ES devices 1 and 6 (Case 1) provide (Case 15). Negative values imply power absorption. Note that the scales on
the best damping, while injections from all six ES devices (Case 4) provide the y-axes are different.
better damping than injections from only ES devices 3 and 4 (Case 15).
8 8