484 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 30, NO.

1, JANUARY 2015

An Adaptive Power Oscillation Damping Controller

by STATCOM With Energy Storage
Mebtu Beza, Student Member, IEEE, and Massimo Bongiorno, Member, IEEE

Abstract—This paper deals with the design of an adaptive power and auxiliary stability enhancement function is possible from
oscillation damping (POD) controller for a static synchronous the energy sources [7]. Because injection of active power is
compensator (STATCOM) equipped with energy storage. This is used temporarily during transient, incorporating the stability en-
achieved using a signal estimation technique based on a modified
recursive least square (RLS) algorithm, which allows a fast, hancement function in systems where active power injection is
selective, and adaptive estimation of the low-frequency electro- primarily used for other purposes [8] could be attractive.
mechanical oscillations from locally measured signals during Low-frequency electromechanical oscillations (typically in
power system disturbances. The proposed method is effective in the range of 0.2 to 2 Hz) are common in the power system
increasing the damping of the system at the frequencies of interest, and are a cause for concern regarding secure system opera-
also in the case of system parameter uncertainties and at various
connection points of the compensator. First, the analysis of the im- tion, especially in a weak transmission system [9]. In this re-
pact of active and reactive power injection into the power system gard, FACTS controllers, both in shunt and series configuration,
will be carried out using a simple two-machine system model. A have been widely used to enhance stability of the power system
control strategy that optimizes active and reactive power injection [1]. In the specific case of shunt connected FACTS controllers
at various connection points of the STATCOM will be derived [STATCOM and static var compensator (SVC)], first swing sta-
using the simplified model. Small-signal analysis of the dynamic
performance of the proposed control strategy will be carried out. bility and POD can be achieved by modulating the voltage at
The effectiveness of the proposed control method to provide power the point of common coupling (PCC) using reactive power in-
oscillation damping irrespective of the connection point of the jection. However, one drawback of the shunt configuration for
device and in the presence of system parameter uncertainties will this kind of applications is that the PCC voltage must be reg-
be verified through simulation and experimental results. ulated within specific limits (typically between 10% of the
Index Terms—Energy storage, low-frequency oscillation, power rated voltage), and this reduces the amount of damping that can
oscillation damping (POD), recursive least square (RLS), static be provided by the compensator. Moreover, the amount of in-
synchronous compensator (STATCOM).
jected reactive power needed to modulate the PCC voltage de-
pends on the short circuit impedance of the grid seen at the con-
I. INTRODUCTION nection point. Injection of active power, on the other hand, af-
fects the PCC-voltage angle (transmission lines are effectively
reactive) without varying the voltage magnitude significantly.
S TATIC synchronous compensator (STATCOM) is a key
device for reinforcement of the stability in an ac power
system. This device has been applied both at distribution level to
The control of STATCOM with energy storage (named here-
after as E-STATCOM) for power system stability enhancement
has been discussed in the literature [10]–[12]. However, the im-
mitigate power quality phenomena and at transmission level for
pact of the location of the E-STATCOM on its dynamic perfor-
voltage control and power oscillation damping (POD) [1]–[3].
mance is typically not treated. When active power injection is
Although typically used for reactive power injection only, by
used for POD, the location of the E-STATCOM has a significant
equipping the STATCOM with an energy storage connected to
impact on its dynamic performance. Moreover, the typical con-
the dc-link of the converter, a more flexible control of the trans-
trol strategy of the device for POD available in the literature is
mission system can be achieved [4], [5]. An installation of a
similar to the one utilized for power system stabilizer (PSS) [9],
STATCOM with energy storage is already found in the U.K.
where a series of wash-out and lead-lag filter links are used to
for power flow management and voltage control [6]. The in-
generate the control input signals. This kind of control strategy
troduction of wind energy and other distributed generation will
is effective only at the operating point where the design of the
pave the way for more energy storage into the power system
filter links is optimized, and its speed of response is limited by
the frequency of the electromechanical oscillations.
In this paper, a control strategy for the E-STATCOM when
Manuscript received November 26, 2013; revised February 18, 2014 and
March 18, 2014; accepted April 23, 2014. Date of publication May 13, 2014; used for POD will be investigated. Thanks to the selected local
date of current version December 18, 2014. This work was supported by signal quantities measured in the system, the control strategy
ELFORSK under the ELEKTRA program. Paper no. TPWRS-01523-2013.
optimizes the injection of active and reactive power to provide
The authors are with the Department of Energy and Environment, Division of
Electric Power Engineering, Chalmers University of Technology, Göteborg 412 uniform damping at various locations in the power system. It
96, Sweden (e-mail: [email protected]; massimo.bongiorno@chalmers. will be shown that the implemented control algorithm is robust
se).
against system parameter uncertainties. For this, a modified re-
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. cursive least square (RLS)-based estimation algorithm as de-
Digital Object Identifier 10.1109/TPWRS.2014.2320411 scribed in [13], [14] will be used to extract the required control

0885-8950 © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
BEZA AND BONGIORNO: ADAPTIVE POWER OSCILLATION DAMPING CONTROLLER BY STATCOM WITH ENERGY STORAGE 485

Fig. 1. Simplified two-machine system with E-STATCOM. Fig. 3. Equivalent circuit for two-machine system with E-STATCOM.

respectively. In the notation in Fig. 2, the superscript “ ” de-

notes the corresponding reference signals.
In this paper, the outer control loop is assumed to be a
POD controller, and the detail of the block will be described
Fig. 2. Block diagram of the control of E-STATCOM.
in Section III. For this reason, we assume that the injected
active and reactive powers in the steady state are zero. When
designing a cascaded controller, the speed of outer control loop
signals from locally measured signals. Finally, the effectiveness
is typically selected to be much slower than the inner one to
of the proposed control strategy will be validated via simulation
guarantee stability. This means that the current controller can
and experimental verification.
be considered infinitely fast when designing the parameters of
the outer controller loop. Therefore, the E-STATCOM can be
II. SYSTEM MODELING FOR CONTROLLER DESIGN modeled as a controlled ideal current source, as depicted in the
A simplified power system model, such as the one depicted equivalent circuit in Fig. 3, for analysis purpose.
in Fig. 1, is used to study the impact of the E-STATCOM on the The level of power oscillation damping provided by the
power system dynamics. The investigated system approximates converter depends on how much the active power output from
an aggregate model of a two-area power system, where each the generators is modulated by the injected current, . For the
area is represented by a synchronous generator. system in Fig. 3, the change in active power output from the
The synchronous generators are modeled as voltage sources generators due to injected active and reactive power from the
of constant magnitude and dynamic rotor angles E-STATCOM is calculated as in
behind a transient reactance . The trans-
mission system consists of two transformers represented by
their equivalent leakage reactance and a transmis-
sion line with equivalent reactance . The
losses in the transmission system are neglected for simpler ana-
lytical expressions. If the mechanical damping in the generators
is neglected, the overall damping for the investigated system (2)
is equal to zero. Therefore, the model is appropriate to allow a
conservative approach of the impact of the E-STATCOM when
used for stability studies [14]. For analysis purpose, the elec- where and represent the
trical connection point of the converter along the transmission change in active power from the corresponding generators due
line is expressed by the parameter as to injected active power and reactive power , re-
spectively. , , and are given by
(1)
where

(3)
The control of the E-STATCOM consists of an outer con-
trol loop and an inner current control loop, as shown in Fig. 2. The initial steady-state PCC voltage magnitude and gener-
The outer control loop, which can be an ac voltage, dc-link ator rotor angles correspond to the operating point
voltage or POD controller, sets the reference current for the where the converter is in idle mode. A derivation to the expres-
inner current controller. The generic measured signal de- sions in (2) is given in the Appendix.
pends on the type of outer loop control. The control algorithm It can be seen from (2) and (3) that the change in active power
is implemented in -reference frame where a phase-locked output from the generators depends on the location of the con-
loop (PLL) [15] is used to track the grid-voltage angle from verter as well as on the amount of injected active and reactive
the grid-voltage vector . By synchronizing the PLL with the power. Moreover, it can be understood from (2) that the effect
grid-voltage vector, the - and -components of the injected cur- of reactive power injection depends on the magnitude and di-
rent ( and ) control the injected active and reactive power, rection of transmitted power from the generators.
486 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 30, NO. 1, JANUARY 2015

III. POD CONTROLLER DESIGN tors due to injected active power is maximum when the com-
The derivation of the POD controller from locally measured pensator is installed at the generator terminals (i.e. and
signals will be made in this section. ). Assuming equal inertia constant for the two generators,
no damping is provided by injection of active power at the elec-
A. Derivation of Control Input Signals trical midpoint of the line (i.e., for ) as
the power output of the two generators is the same and the net
Considering the simplified two-machine system in Fig. 1,
impact is zero. At this location, the derivative of PCC-voltage
the active power output from each generator should change in
phase is zero [see (6)]. This means that scales the speed
proportion to the change in its speed to provide damping [9].
variation of the two generators depending on the location of
From (2), it can be observed that the effect of the power in-
E-STATCOM and its magnitude changes in proportion to the
jected by the compensator on the generator active power output
level of damping by active power injection. Therefore,
highly depends on the parameter , i.e., on the location of the
is an appropriate input signal for controlling the active power
E-STATCOM. Using the equivalent system in Fig. 3, a control
injection. On the other hand, it can be understood from (2) that
input signal that contains information on the speed variation of
the change in the output power from the generators due to in-
the generators can be derived. When the E-STATCOM is not
jected reactive power is maximum at the electrical midpoint of
injecting any current, the variation of the locally measured sig-
the line (i.e., ) and minimum at the generator terminals
nals, and at different E-STATCOM connection points
(i.e., and ). As the changes in the power output
using the dynamic generator rotor angles and is given
of the two generators are the same in magnitude and opposite
by
in sign, a signal that varies linearly with the speed variation be-
tween the two generators, is an appropriate signal to con-
(4)
trol reactive power injection. This information can be obtained
from the derivative of the transmitted active power .
(5)
B. Estimation of Control Input Signals
From a small-signal point of view and under the assump- As described in the Introduction, effective power oscilla-
tion that the PCC-voltage magnitude along the line does tion damping for various power system operating points and
not change significantly, the required control input signals can E-STATCOM locations require fast, accurate, and adaptive es-
be derived from the PCC-voltage phase and transmitted active timation of the critical power oscillation frequency component.
power as [14] This is achieved by the use of an estimation method based on a
modified RLS algorithm. For reasons described in the previous
(6) subsection, the derivative of the PCC-voltage phase and the
transmitted power should be estimated for controlling the ac-
tive and reactive power injection, respectively. The aim of the
(7) algorithm is therefore to estimate the signal components that
consist of only the low-frequency electromechanical oscillation
where the constant has been defined in the previous section. in the measured signals and . By using a PLL with
The nominal system frequency is represented by whereas bandwidth much higher than the frequency of electromechan-
and represent the speed variation of the generators ical oscillations, the derivative of the PCC-voltage phase can
in p.u. The electromechanical dynamics for each generator be obtained from the change in frequency estimate of the
is given by [9] PLL . Therefore, the low-frequency electro-
mechanical oscillation component can be extracted directly
(8) from the frequency estimate of the PLL. On the other hand, the
derivative of transmitted power is estimated by extracting the
where , , , , and represent inertia con- low-frequency electromechanical oscillation component from
stant, speed variation, change in input torque, change in output the measured signal, and then applying a phase shift of
torque and mechanical damping constant for the th generator, to the estimated oscillation frequency component.
respectively. From the estimated control input signals
The derivative of the PCC-voltage phase and transmitted ac- and , which contain only a particular oscillation
tive power are both dependent on the speed variation of the gen- frequency component, the reference injected active and reac-
erators. Moreover, the derivative of the PCC-voltage phase de- tive current components from the E-STATCOM can
pends on the location of E-STATCOM, through the parameter be calculated to setup the POD controller as in Fig. 4. The terms
, as well as the mechanical dynamics of the generators as and represent proportional controller gains for the ac-
shown in (8). This information will be exploited in the POD tive and reactive current components, respectively.
controller design. To describe the estimation algorithm, an input signal which
For the two machine system in Fig. 1, damping is related to could be either or , as shown in Fig. 4, is considered.
the variation of the speed difference between the two genera- Following a power system disturbance, will consist of an av-
tors, . From (2) and (3), it can be un- erage value that varies slowly and a number of low-frequency
derstood that the change in the output power from the genera- oscillatory components, depending on the number of modes that
BEZA AND BONGIORNO: ADAPTIVE POWER OSCILLATION DAMPING CONTROLLER BY STATCOM WITH ENERGY STORAGE 487

Fig. 4. Block diagram of the POD controller.

are excited by the disturbance. For simplicity, let us assume that

there exist a single oscillatory component in the input signal.
Therefore, the input signal consists of an average component
and an oscillatory component which can be modeled Fig. 5. Block diagram of the modified RLS estimator for multiple oscillation
modes.
as

(9)
frequency selectivity of the algorithm reduces. For this reason,
where is expressed in terms of its amplitude , fre- the conventional RLS algorithm must be modified in order
quency and phase . The model in (9) is rewritten using to achieve fast transient estimation without compromising its
the oscillation angle as steady-state selectivity. In this paper, this is achieved with the
use of variable forgetting factor as described in [13]. When the
RLS algorithm is in steady-state, its bandwidth is determined
by the steady-state forgetting factor . If a rapid change is
where the terms and are given by detected in the input (i.e., if the estimation error magnitude,
exceeds a predefined threshold), will be modified
to a smaller transient forgetting factor . Thus, by using
a high-pass filter with time constant , will be slowly
From an observation matrix and measured input signal increased back to its steady-state value .
, the estimated state vector is derived using the RLS al- Besides , the performance of the estimation method depends
gorithm in discrete time as [13], [14] on accurate knowledge of the oscillating frequency, . This
frequency is dependent on the system parameters and its oper-
(10) ating conditions. If the frequency content of the input changes,
the estimator will give rise to a phase and amplitude error in the
with estimated quantities. Therefore, a frequency adaptation mecha-
nism as described in [14] is implemented to track the true oscil-
lation frequency of the input from the estimate of the oscillatory
component, .
Modification for Multiple Oscillation Modes: The investi-
Calling the identity matrix, the gain matrix and covari- gated control method has been derived under the assumption of
ance matrix are calculated recursively starting with an initial a single oscillatory frequency component in the input signal. A
invertible matrix as brief description of how the proposed algorithm can be extended
for multi-area system with multiple oscillation modes will be
briefly presented here for future reference. Assuming that the
(11) input signal contains oscillatory components, (9) must be
modified as
(12)

where represents the forgetting factor for the RLS algorithm

such that . With representing the sampling time,
the steady-state bandwidth of the RLS and the estimation
error are given by [14]
(13)

Modification in the Conventional RLS Algorithm: The se- where the th oscillation mode (with ) is
lection of is a tradeoff between a good selectivity for the expressed in terms of its amplitude , frequency ,
estimator and its speed of response [13], [14]. A high forgetting and phase . Using the model in (13), the RLS described in
factor results in low estimation speed with good frequency se- the previous sections (including variable forgetting factor and
lectivity. With increasing estimation speed (decreasing ), the frequency adaptation for each considered oscillation mode) can
488 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 30, NO. 1, JANUARY 2015

be modified as described in Fig. 5. Thus, the POD controller in

Fig. 4 can be modified accordingly to control each mode inde-
pendently. Observe that the phase-shift applied for calculation
of the reference currents depends on the investigated system and
needs to be calculated for each oscillatory mode [9].

IV. STABILITY ANALYSIS OF SYSTEM MODEL

The mathematical model of the system in Fig. 3 is developed
in this section to investigate the performance of the POD con-
troller using active and reactive power injection. Using the ex-
pressions in (6)–(7) for and , the injected cur-
rents are controlled as

(14) (a) (b) (c)

Fig. 6. Real and imaginary part of the complex conjugate poles versus position.
(a) Active power injection. (b) Reactive power injection. (c) Active and reactive
(15) power injection. 0.4444 p.u. .

where the constant is as defined in (3). Linearizing around

an initial steady-state operating point, the small-signal dynamic sign of should be chosen based on the sign of . For a
model of the two-machine system with the E-STATCOM in per transmitted power from Generator 1 to Generator 2, will be
unit is developed as in positive and the sign of should be negative. For a trans-
mitted power in the other direction, the sign of should be
opposite. The terms and are the cross coupling terms
between the two generator speed variations.
With active power injection only , the cross cou-
pling terms reduce the damping as the speed variation of the
(16)
generators will be opposite at the oscillatory frequency. At the
mass-scaled electrical midpoint of the line where ,
where represents the rotor angle dif- the damping that can be provided by is zero. Therefore,
ference between the two generators and other signals as defined the active power injected by the E-STATCOM at this location is
previously. Assuming no mechanical damping and the initial set to zero by the control algorithm. When moving away from
steady-state speed of the generators set to , the constants are this point towards the generator terminals, increases and at
derived as in the same time the cross coupling terms decrease. This enhances
the damping that can be provided by active power injection and
therefore the amount of injected active power is increased. In
the case of reactive power injection only , positive
damping is provided by the cross coupling terms and maximum
damping is provided at the electrical midpoint of the line (i.e.,
(17) for a symmetrical system) where magnitude is
maximum.
As an example for the analysis in this section, a hypothet-
ical 20/230 kV, 900 MVA transmission system similar to the
one in Fig. 1 with a total series reactance of 1.665 p.u. and in-
where is given by ertia constant of the generators s is con-
sidered. The leakage reactance of the transformers and transient
impedance of the generators are 0.15 p.u. and 0.3 p.u., respec-
tively. The movement of the poles for the system as a function of
(18) the E-STATCOM location is shown in Fig. 6. With the described
control strategy, injected active power is zero at the point where
The terms and represent the synchronizing torque the effect of active power injection on damping is zero. This is at
coefficients resulting from the selected operating point and the the electrical midpoint of the line. On the other hand, at the same
contribution of the E-STATCOM is zero. The terms and location damping by reactive power injection is maximum. The
determine the damping torque coefficient provided by the reverse happens at either end of the generators. Thanks to a good
E-STATCOM with respect to the change in speed of the respec- control of and , it is also possible to see from Fig. 6 that
tive generator. To provide positive damping, and should a more uniform damping along the line is obtained by using in-
be negative. For this, the sign of should be negative and the jection of both active and reactive power.
BEZA AND BONGIORNO: ADAPTIVE POWER OSCILLATION DAMPING CONTROLLER BY STATCOM WITH ENERGY STORAGE 489

Fig. 7. Simplified two-area four machine power system.

Fig. 9. Single-line diagram of the laboratory setup.

(a) (b)

Fig. 10. Photograph of the laboratory setup. (a) Analog transmission-line

model. (b) Generator system.

is achieved by active power injection (see Fig. 8, black solid

plots). With respect to reactive power injection, maximum
damping action is provided when the E-STATCOM is con-
nected close to the electrical midpoint of the line and the level
of damping decreases when moving away from it (see Fig. 8,
gray solid plots). Because of a good choice of signals for
controlling both active and reactive power injection, effective
Fig. 8. Measured transmitted active power output following a three-phase fault
with E-STATCOM connected at bus 7 (top), bus 8 (middle), and bus 9 (bottom). power oscillation damping is provided by the E-STATCOM
POD by only (black solid), only (gray solid), both , (black irrespective of its location in the line (see Fig. 8, black dashed
dashed), and no POD (gray dashed). plots).

VI. EXPERIMENTAL VERIFICATION

V. SIMULATION RESULTS
To validate the results obtained via simulation for the POD
The POD controller described in Section III is here verified
controller performance by E-STATCOM, the obtained experi-
via PSCAD/EMTDC simulation using the well known two-area
mental results will be presented in this section.
four-machine system in Fig. 7. The implemented system is rated
20/230 kV, 900 MVA and the parameters for the generators and
A. Laboratory Setup
transmission system together with the loading of the system are
given in detail in [9]. The system is initially operating in steady- The schematic of the laboratory setup is shown in Fig. 9,
state with a transmitted active power, 400 MW from while Fig. 10 shows a photograph of the actual transmission
area 1 to area 2. A three-phase fault is applied to the system on system model. The system consists of a 75-kVA, 400-V syn-
one of the transmission lines between bus 7 and bus 8. The fault chronous generator connected to a stiff ac grid through a trans-
is cleared after 120 ms by disconnecting the faulted line. Due to mission-line model. A voltage source converter (VSC) system
the applied disturbance, a poorly damped oscillation is obtained can be connected at various locations of the transmission line.
after the fault clearing. 1) VSC System: The VSC system consists of a two-level con-
With the POD controller structure described in Fig. 4, the verter connected to the transmission line model though a series
performance of the E-STATCOM following the fault at three reactor with 2 mH, 6.2 m and shunt capacitor
different locations is shown in Fig. 8. As described in the with 60 uF. The VSC is controlled from a computer with
small-signal analysis for two-machine system in Section IV, a dSpace 1103 board, which can be programmed using C-code
when moving closer to the generator units, a better damping or MATLAB/Simulink. The dc-link of the VSC is connected to a
490 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 30, NO. 1, JANUARY 2015

Fig. 11. Measured transmitted active power from the generator. A three-phase
fault is applied and cleared after 250 ms.

dc machine rated 700 V and 60 A. The dc machine is equipped Fig. 12. Modified RLS-based estimator. Variation of forgetting factor (top)
with field control and the terminal dc voltage is controlled to and estimate of oscillation frequency (bottom).
650 V for all of the experiments. The dc machine will act as the
energy source providing active power injection capability to the
VSC. Therefore, the VSC with the dc machine is considered as
an E-STATCOM.
2) Network Model: The network model is a down-scaled
version of an actual Swedish 400 kV transmission system
with the model rated at 400 V, 50 Hz. The transmission-line
model consists of six identical sections (with parameters
2.05 mH, and 46 F), each
corresponding to a portion of 150 km of the actual line [16].
As shown in Fig. 9, faults can be applied to the transmission
system using the contactor (CT1).
3) Synchronous Generator: A photograph of the gener-
ator system which represents an actual model of the large
Harsprånget hydro power plant, situated by the Luleå river
in northern Sweden is shown in Fig. 10. The synchronous
generator has an inertia constant of 5.6 s, and it is driven by an
85-kW dc motor [16].

B. Experimental Verification of Modified RLS Estimator

In order to create a power oscillation in the system, a
three-phase fault has been applied in the middle of the trans-
mission line. The fault clearing time is 250 ms. The measured
output power from the generator as shown in Fig. 11 is
then used as an input to test the performance of the algorithm
to estimate the low-frequency electromechanical oscillation Fig. 13. Estimate for the average and oscillatory component. Top: estimated
component. The actual oscillation frequency of the measured average component. Middle: estimated oscillatory component. Bottom: mea-
sured generator active power signal (gray line) and estimated signal (black line).
signal is about 0.42 Hz. This low oscillation frequency
highlights the importance of the adopted estimation method,
since the classical approaches based on filters would require
low bandwidth, resulting in a reduction in the estimation speed thanks to the variation of , very fast transient performance is
[14]. To investigate the robustness of the algorithm against achieved. The frequency adaptation ensures a correct extraction
system parameter changes, an initial oscillatory frequency of of the magnitude and phase of the oscillatory component which
0.6 Hz is assumed in the estimation. is a crucial point in the POD application.
The modified RLS algorithm as described in Section III-B
C. Experimental Verification of POD Controller
uses a variable forgetting factor and a frequency adaptation
by E-STATCOM
mechanism. The trend of the variable forgetting factor and fre-
quency adaptation is shown in Fig. 12. The performance of the As described in Section III, the outputs of the POD controller
algorithm to estimate the average and oscillatory components are active and reactive current references to be injected to the
of the measured signal is shown in Fig. 13. As the results show, grid by the E-STATCOM. The current controller implemented
BEZA AND BONGIORNO: ADAPTIVE POWER OSCILLATION DAMPING CONTROLLER BY STATCOM WITH ENERGY STORAGE 491

Fig. 14. Single-line diagram of the laboratory setup for POD.

Fig. 16. Injected active and reactive power with E-STATCOM connected at bus
1 (black line) and bus 2 (gray line). Active power injection (top) and reactive
power injection (bottom); both and used for POD.

the distance from the generator where its impact to provide

damping decreases (see Fig. 16). A better damping with active
power injection is obtained when the E-STATCOM is closer to
the generator, in this case at bus 1 (see Fig. 15). With respect
to reactive power injection, the damping provided by the com-
pensator increases when moving closer to the electrical mid-
Fig. 15. Measured generator active power output following a three-phase fault point. As the transmitted power is used to control reactive
with E-STATCOM connected at bus 1 (top) and bus 2 (bottom). POD by
only (black solid line), only (gray solid line), both , (black dashed power, the same amount of injected reactive power is used at
line) and no POD (gray dashed line). the two locations and a better damping action is provided close
to the electrical midpoint of the line, in this case at bus 2. With a
proper choice of control signals for injection of active and reac-
here is a Proportional-Integral based vector controller. It is de- tive power, effective power oscillation damping is provided by
signed based on the internal model control approach. The band- the E-STATCOM at both connection points of the compensator
width of the controller is chosen as 2500 rad/s, and the detail as shown in Fig. 15 (black dashed curves).
of the implementation can be found in [14]. To verify the POD To test the dynamic performance of the investigated POD
controller performance, the setup in Fig. 9 which represents a controller in case of system parameter changes, a second set of
single-machine infinite bus system with E-STATCOM is used. experiments has been carried out assuming an oscillation fre-
The single-line diagram of the setup is shown in Fig. 14, where quency of 0.9 Hz, where the actual measured oscillation fre-
the possible connection buses of the E-STATCOM are marked quency is 0.42 Hz. This means that an error of 100% in the es-
as 1 and 2. For the tests, the gain is chosen to get a damping timated oscillation frequency is here considered. Fig. 17 com-
ratio of 10% at bus 2 when reactive power is used for POD. For pares the performance of the POD controller with and without
a fair comparison, the gain is then adjusted to get a similar the oscillation frequency adaptation in the RLS estimator. In
order of maximum active and reactive power injection for the both cases, the E-STATCOM is connected at bus 2 and injec-
tests. Once the values for the gains are selected, they are kept tion of active and reactive power is used for POD. By using the
constant for all the experiments. frequency adaption as described in Section III-B, the phase of
First, the power oscillation damping controller is tested with the oscillatory component in the input signal is correctly esti-
two connection points of the E-STATCOM as shown in Fig. 14. mated, thus providing an effective damping. This is advanta-
For this test, knowledge of the oscillatory frequency in the trans- geous when compared to the classical approaches, where the
mitted active power has been considered. As before, a three- correct phase shift is provided in the estimation only at a par-
phase fault is applied at bus 1 and the fault is cleared after ticular oscillation frequency. As shown in Fig. 17 (gray solid
250 ms. The performance of the E-STATCOM for POD using plots), if the POD controller is not adapted to changes in the
the control strategy described in Section III is shown in Figs. 15 system, its performance is significantly reduced. This is also
and 16. Observe that, to facilitate the comparison, the presented shown in Fig. 18, where the total energy exchange be-
measured signals have been filtered to remove noise and high- tween the E-STATCOM and the grid until the oscillations are
frequency harmonic components. A low-pass filter with cutoff completely damped for the two cases is displayed. The case
frequency of 50 Hz has been used in order not to affect the without frequency adaptation is characterized by a longer set-
low-frequency components. As described in the small-signal tling time of the oscillation and a larger amount of total energy
analysis in Section IV, the injected active power decreases with exchange. This results in an uneconomical use of the energy
492 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 30, NO. 1, JANUARY 2015

VII. CONCLUSION
An adaptive POD controller by E-STATCOM has been
developed in this paper. For this, a modified RLS algorithm has
been used for estimation of the low-frequency electromechan-
ical oscillation components from locally measured signals
during power system disturbances. The estimator enables a
fast, selective and adaptive estimation of signal components at
the power oscillation frequency. The dynamic performance of
the POD controller to provide effective damping at various con-
nection points of the E-STATCOM has been validated through
simulation as well as experimental verification. The robustness
of the control algorithm against system parameter changes
has also been proven through experimental tests. Furthermore,
using the frequency variation at the E-STATCOM connection
point as the input signal for the active power modulation, it has
been shown that active power injection is minimized at points
in the power system where its impact on POD is negligible.
This results in an optimal use of the available energy source.

APPENDIX
Here, the relation between the change in active power output
from the generators and the active and reactive power injected
by the E-STATCOM will be derived. With the notations given
Fig. 17. Top: measured generator output power with frequency adaptation in Fig. 3, the power output of Generator 1 is given by
(black solid), without frequency adaptation (gray solid line) and with no POD
(gray dashed line). Middle: injected active power with frequency adaptation
(black line) and without frequency adaptation (gray line). Bottom: injected (19)
reactive power with frequency adaptation (black) and without frequency
adaptation (gray line).
From a small-signal point of view, the change in PCC voltage
from its initial steady-state value due to injection
of active and reactive current is given by

(20)

where represents the Thevenin equivalent impedance at the

PCC. Equation (20) can be expressed in terms of amplitude and
phase change as

(21)

The change in active power output of Generator 1 due to in-

jection of active and reactive current from the E-STATCOM is
derived from (19)–(21) as

(22)
Fig. 18. Total energy exchange to damp oscillations with and without fre-
quency adaptation. (23)

In steady state, i.e., when the E-STATCOM is in idle mode,

the PCC voltage can be expressed as
storage. Observe that in order to avoid continuous injection of
active and reactive power, the POD controller implemented in (24)
the laboratory sets the reference active and reactive currents to
zero when the amplitude of the oscillatory signal goes below a where is the relative electrical location of the E-STATCOM
pre-defined threshold. Here, this threshold is set to 0.005 pu. As as defined in (1).
an example, it is clear from Fig. 17 (middle and bottom plots) Substituting (24) into (22) and (23) and defining , and
that, when the adaptive POD controller is used, the amplitude as in (3), the change in active power output from Gener-
of the power oscillation crosses this threshold at 8 s. As a result, ator 1 due to injection of active and reactive power from the
the active and reactrive powers injected by the compensator be- E-STATCOM given in (2) can be derived. By conservation of
come zero. energy, the expression for the change in active power output
BEZA AND BONGIORNO: ADAPTIVE POWER OSCILLATION DAMPING CONTROLLER BY STATCOM WITH ENERGY STORAGE 493

from Generator 2 due to injection of active and reactive power [12] A. Arsoy, L. Yilu, P. F. Ribeiro, and F. Wang, “Power converter and
is calculated as and SMES in controlling power system dynamics,” in Proc. Ind. Appl.
Conf., Oct. 2000, vol. 4, pp. 2051–2057.
. [13] M. Beza and M. Bongiorno, “A fast estimation algorithm for low-fre-
quency oscillations in power systems,” in Proc. 14th Eur. Conf. Power
Electron. Appl., Sep. 2011, pp. 1–10.
[14] M. Beza, “Control of energy storage equipped shunt-connected con-
REFERENCES verter for electric power system stability enhancement,” Licentiate
Thesis, Dept. Energy and Environment, Chalmers Univ. of Technol.,
[1] N. G. Hingorani and L. Gyugyi, Understanding FACTS. Concepts and Gothenburg, Sweden, 2012, .
Technology of Flexible AC Transmission Systems. New York, NY, [15] L. Ängquist and M. Bongiorno, “Auto-normalizing phase-locked loop
USA: IEEE, 2000. for grid-connected converters,” in Proc. IEEE Energy Conv. Congress
[2] G. Cao, Z. Y. Dong, Y. Wang, P. Zhang, and Y. T. Oh, “VSC based Expo., Sep. 2009, pp. 2957–2964.
STATCOM controller for damping multi-mode oscillations,” in Proc. [16] M. Gustafsson and N. Krantz, “Voltage collapse in power systems,” Li-
IEEE Power and Energy Soc. General Meeting—Conversion and De- centiate thesis, Dept. Electr. Power Eng., Chalmers Univ. of Technol.,
livery of Electrical Energy in the 21st Century, Jul. 2008, pp. 1–8. Gothenburg, Sweden, Dec. 1995.
[3] M. Zarghami and M. L. Crow, “Damping inter-area oscillations in
power systems by STATCOMs,” in Proc. 40th North Amer. Power Mebtu Beza (S’13) received the M.Sc. and Lic.
Symp., Sep. 2008, pp. 1–6. Eng. degrees in electrical power engineering from
[4] Z. Yang, C. Shen, L. Zhang, M. L. Crow, and S. Atcitty, “Integration of Chalmers University of Technology, Göteborg,
a statcom and battery energy storage,” IEEE Trans. Power Syst., vol. Sweden, in 2009 and 2012, respectively, where he is
16, no. 2, pp. 254–260, May 2001. currently working toward the Ph.D. degree.
[5] A. Arulampalam, J. B. Ekanayake, and N. Jenkins, “Application study His research interests include signal processing
of a STATCOM with energy storage,” Proc. Inst. Electr. Eng.—Gener., in power systems, control of grid-connected con-
Transm. and Distrib., vol. 150, pp. 373–384, July 2003. verters, and application of power electronics in
[6] N. Wade, P. Taylor, P. Lang, and J. Svensson, “Energy storage for power systems.
power flow management and voltage control on an 11 kV UK distribu-
tion network,” Prague, Czech Republic, CIRED paper 0824, Jun. 2009.
[7] A. Adamczyk, R. Teodorescu, and P. Rodriguez, “Control of full-scale
converter based wind power plants for damping of low frequency
system oscillations,” in Proc. IEEE PowerTech, Trondheim, Norway,
Jun. 2011, pp. 1–7. Massimo Bongiorno (S’02–M’07) received the
[8] H. Xie, “On power-system benefits, main-circuit design, control of M.Sc. degree in electrical engineering from the
Statcoms with energy storage,” Ph.D. dissertation, Dept. Electr. En- University of Palermo, Palermo, Italy, in 2002,
ergy Conversion, Royal Inst. Technol., Stockholm, Sweden, 2009. and the Lic.Eng and Ph.D. degrees from Chalmers
[9] P. Kundur, Power System Stability and Control. New York, NY, University of Technology, Göteborg, Sweden, in
USA: McGraw-Hill, 1994. 2004 and 2007, respectively.
[10] K. Kobayashi, M. Goto, K. Wu, Y. Yokomizu, and T. Matsumura, From 2007 to 2010, he was an Assistant Professor
“Power system stability improvement by energy storage type with the Department of Energy and Environment,
STATCOM,” in Proc. IEEE Power Tech Conf., Bologna, Italy, Jun. Chalmers University of Technology, Göteborg,
2003, vol. 2, DOI 10.1109/PTC.2003.1304302. Sweden, where he became an Associate Professor in
[11] L. Zhang and Y. Liu, “Bulk power system low frequency oscillation power electronics applications for power systems in
suppression by FACTS/ESS,” in Proc. IEEE PES Power Syst. Conf. September 2010. His research interests include application of power electronics
Exp., Oct. 2004, pp. 219–226. in power systems and power quality.