power oscillation damping control by pss and dfig wind...
Post on 12-Mar-2018
215 Views
Preview:
TRANSCRIPT
Turk J Elec Eng & Comp Sci
(2017) 25: 4354 – 4368
c⃝ TUBITAK
doi:10.3906/elk-1702-190
Turkish Journal of Electrical Engineering & Computer Sciences
http :// journa l s . tub i tak .gov . t r/e lektr ik/
Research Article
Power oscillation damping control by PSS and DFIG wind turbine under multiple
operating conditions
Korakot THANPISIT∗, Issarachai NGAMROODepartment of Electrical Engineering, Faculty of Engineering, King Mongkut’s Institute of Technology Ladkrabang,
Bangkok, Thailand
Received: 15.02.2017 • Accepted/Published Online: 08.06.2017 • Final Version: 05.10.2017
Abstract:Multiple operating conditions in power systems including wind power sources significantly affect the damping
of low frequency oscillation modes due to diverse generating and loading conditions, random wind speeds, line outage
contingencies, etc. To cope with multiple operating conditions, this paper proposes the new parameter optimization
technique of the power system stabilizer (PSS) and the doubly-fed induction generator (DFIG) wind turbine with the
power oscillation damper (POD) based on the probability method. Different operating conditions are randomly generated
by Monte Carlo simulation. Under the generated operating points, the particle swarm optimization of PSS and POD
parameters is carried out to achieve the highest probability that the damping ratios of all oscillation modes are greater
than the desired damping ratio for all operating points. Study results in the IEEE New England 39-bus system indicate
that under the occurrence of faults, the PSS and POD optimized by the proposed method yield better stabilizing
performance than the conventional PSS and POD over a wide range of operating points.
Key words: Doubly-fed induction generator, power oscillation damper, probability method, power system stabilizer,
Monte Carlo simulation
1. Introduction
Generally, the damping of power oscillations with low frequency oscillation between 0.1 and 2.0 Hz, i.e. local and
interarea oscillations, is influenced by several loading and generating conditions, line outage contingencies, etc.
[1]. Without effective countermeasures, undamped power oscillations may lead to an unstable system and result
in a wide-area blackout. To suppress power oscillations, power system stabilizers (PSSs), which are the most
cost-effective devices, have been used for a long time. A PSS is equipped with the automatic voltage regulator
(AVR) of a synchronous generator. By adding a stabilizing signal to the AVR, the PSS can be effectively applied
to solve the problem of local and interarea oscillations in conventional power systems [1].
Nowadays, the penetration of wind power generation widely increases in power systems due to being
abundant, sustainable, and environmentally friendly [2]. By the end of 2015, the global cumulative installed
wind capacity was about 430 GW [3]. Nevertheless, the intermittent power output from wind power sources
makes the power oscillations problem more complicated [4]. Under these situations, the PSS may no longer be
able to suppress power oscillations.
Among wind turbines, the doubly-fed induction generator (DFIG) is popularly adopted in wind energy
conversion systems. The DFIG wind turbine can generate maximum power with variable wind speeds. Further-
∗Correspondence: silver brema@hotmail.com
4354
THANPISIT and NGAMROO/Turk J Elec Eng & Comp Sci
more, it has many advantages such as low power rating of the converter, low loss, and high efficiency [5]. The
salient feature of the DFIG is the ability of active and reactive power output controls. In particular, the active
and reactive power outputs can be controlled to eliminate power oscillations by installing a power oscillation
damper (POD) in the speed and voltage control loops, respectively [6]. The comparison of power oscillation
damping effect between the active and reactive power controls by DFIG wind turbine was studied in [7]. Study
results signified that the active power control may invoke the torsional oscillation of the wind turbine. On
the contrary, the reactive power control is immune to the torsional oscillation and provides satisfying damping
effects.
Previous works used the reactive power control of DFIG wind turbines to damp power oscillations. A POD
with second-order lead/lag compensator structure was designed by the root locus method [7] and particle swarm
optimization [8] so that the damping ratios of oscillation modes could be enhanced. In [9], the optimal tuning of
a POD using mixed H2/H∞ control was presented to obtain the robustness of the POD against various system
operations. In [10], a robust decentralized POD design to improve the system’s robust stability margin by inverse
multiplicative perturbation was proposed. Nevertheless, the design methods in [7–10] were carried out under
set operating conditions. They cannot guarantee the damping performance of the designed PODs under other
operating conditions. In addition, to achieve more stabilizing performance, the coordinated control of the PSS
and DFIG with POD is required. In [11], the design technique for coordination between a PSS and DFIG with a
POD considering random wind power output variation was presented by an extended probabilistic small-signal
stability analysis. Study results indicated that the coordinated PSS and POD yield a higher damping effect
than either the PSS or POD alone. The coordinated damping controller can ensure the damping performance
under diverse operating conditions. To enhance the robustness against system uncertainties, the parameters of
a coordinated PSS and POD in [12] were optimized under several operating points by concurrently and equally
increasing the output power of synchronous generators.
To deal with multiple operating conditions due to various load demands, power generations from syn-
chronous generators, random wind speeds, and line outage contingencies, this paper focuses on the new pa-
rameter optimization of a PSS and DFIG wind turbine with a POD based on the probability method. The
PSS and POD parameters are optimized for the probability that the damping ratios of all oscillation modes
are greater than the desired damping ratios for all random events generated by Monte Carlo simulation (MCS).
The optimization problem is solved by particle swarm optimization (PSO) so that the optimal parameters of
the PSS and POD can be obtained. The damping performance of the proposed PSS and POD in comparison
with the conventional PSS and POD is carried out in the IEEE 39-bus New England system.
The organization of this paper is described as follows. Section 2 presents the study system and modeling.
Next, Section 3 explains the proposed optimization technique. Subsequently, Section 4 provides simulation
results. Finally, the conclusion is given.
2. Study system and modeling
2.1. Study system
Figure 1 displays the modified IEEE 39-bus New England system [13]. All synchronous generators (G1–G10),
which are installed with turbine governor type II and AVR type III, are represented by the fourth-order model
[14]. After the eigenvalue analysis, nine oscillation modes with percent damping ratio (ζ) and oscillation
frequency (f) are given in Table 1. In this study, the acceptable percent damping ratio of each oscillation mode
4355
THANPISIT and NGAMROO/Turk J Elec Eng & Comp Sci
is at least 5%. In Table 1, it can be observed that the damping ratios of oscillation modes no. 1, 4, 5, 6, 7, 8,
and 9 are less than 5%.
G10
G1
G8
G2 G3 G4 G5
G7
G6
G9
39
9
8
7
5
6
11
3132
10
13
12
14
4
15
16
2122
35
3433
19
20
3
1
2
30 25
37
26 28
27
29
38
24
23
18 17
36
W3
W2
W1
Figure 1. Modified IEEE 39-bus New England system.
Table 1. Eigenvalue analysis results.
Mode Eigenvalues Damping ratio (ζ) Frequency (f)1st –0.3858 ± j9.4495 4.08% 1.50 Hz2nd –0.6649 ± j9.5230 6.96% 1.52 Hz3rd –0.5154 ± j9.2602 5.56% 1.47 Hz4th –0.3059 ± j8.0050 3.82% 1.27 Hz5th –0.3180 ± j7.6725 4.14% 1.22 Hz6th –0.3038 ± j7.0551 4.30% 1.12 Hz7th –0.2041 ± j6.0972 4.57% 0.97 Hz8th –0.2802 ± j6.0219 4.65% 0.96 Hz9th –0.2023 ± j4.0893 4.94% 0.65 Hz
The selection of proper locations of PSSs is evaluated by the sensitivity of PSS effect (SPE) method [15].
A high value of SPE implies a suitable synchronous generator for installing a PSS. The SPE can be calculated
by:
SPE = φ∆ωψ∆vfKe
Te, (1)
where ϕ∆ω is the right eigenvector corresponding to the speed deviation of the synchronous generator, ψ∆vf is
the left eigenvector corresponding to the field voltage deviation, and Ke and Te are the gain and time constant
of the AVR, respectively. As a result, Table 2 shows that the synchronous generators for suitable installation
of PSSs are at G2, G8, and G9.
4356
THANPISIT and NGAMROO/Turk J Elec Eng & Comp Sci
Table 2. SPE analysis results.
G1 G2 G3 G4 G5 G6 G7 G8 G9
SPE 0.0023 0.1134 0.0251 0.0198 0.0340 0.0175 0.0572 0.1067 0.1366
To serve increased load demands, three units of DFIG wind turbines (W1–W3), each with a capacity of
100 MVA, are additionally installed in the study system. The DFIG data are given in the Appendix. A POD
is installed into each DFIG. The aim of the PSS and POD is to ameliorate the damping ratios of all oscillation
modes.
2.2. DFIG model
The structure of the DFIG wind turbine, which is delineated in Figure 2, mainly consists of a grid-side converter
(GSC) and rotor-side converter (RSC). Study results in [16] indicated that damping control using the RSC is
superior to that of the GSC. In this work, the voltage controller in the RSC is used to modulate the reactive
power output of the DFIG to damp out the power oscillations.
Speed
controller
Voltage
controller
PWM
Grid
Induction
machine
Gear boxWind
patterns
Rotor side
converter
Grid side
converter
Figure 2. Structure of DFIG wind turbine and control system.
Here, the reduced fourth-order model of the DFIG [14] is used in the wind turbine-generation system as
the stator and rotor flux dynamics are fast in comparison with grid dynamics and the converter controls. The
steady-state electrical equations of the DFIG are given by
vds = −rsids + ((xs + xµ)iqs + xµiqr)
vqs = −rsiqs − ((xs + xµ)ids + xµidr)
vdr = −rridr + (1− ωm)((xs + xµ)iqr + xµiqs)
vqr = −rriqr − (1− ωm)((xs + xµ)idr + xµids)
, (2)
where vds , vqs , vdr , and vqr are d and q stator and rotor voltages; ids , iqs , idr , and iqr are d and q stator and
rotor currents; rs and rr are stator and rotor resistances; xs is stator reactance; xu is magnetizing reactance;
and ωm is rotor speed.
4357
THANPISIT and NGAMROO/Turk J Elec Eng & Comp Sci
The stator voltages are functions of the grid voltage magnitude and phase as
vds = −v sin(θ)vqs = v cos(θ)
, (3)
where v is a magnitude of DFIG terminal voltage and θ is a phase of terminal voltage.
The generator active and reactive powers depend on the stator and converter currents, as follows:
PDFIG = vdsids + vqsiqs + vdcidc + vqciqc
QDFIG = vqsids − vdsiqs + vqcidc − vdciqc, (4)
where PDFIG and QDFIG are active and reactive power outputs, vdc and vqc are d and q converter voltages,
and idc and iqc are d and q converter currents.
Due to the converter operation mode, the converter powers injected in the grid can be written as a
function of stator and rotor currents as
pc = vdcidc + vqciqc
qc = vqcidc − vdciqc, (5)
where pc and qc are converter active and reactive powers on the grid.
On the rotor side we have
pr = vdridr + vqriqr
qr = vqridr − vdriqr, (6)
where pr and qr are rotor-side active and reactive powers.
Assuming a lossless converter model, the active power of the converter coincides with the rotor active
power and thus pc = pr . The reactive power injected into the grid can be approximated neglecting stator
resistance and assuming that the d-axis coincides with the maximum of the stator flux. The flux-based rotating
reference frame is used to model the DFIG. Due to the initial value of θ = 0, Eq. (3) can be written as
vds = vdr = 0
vqs = vqr = v, (7)
and then Eq. (4) can be written as
PDFIG = v(iqs + iqr)
QDFIG = −v(ids + idr). (8)
The relation between stator fluxes and generator currents is
ψds = −((xs + xµ)ids + xµidr)
ψqs = −((xs + xµ)iqs + xµiqr), (9)
where ψds and ψqs are the d and q axis of stator fluxes, respectively.
Substituting Eq. (9) into Eq. (8), the active and reactive power outputs of the DFIG wind turbine are
represented as the function of iqr and idr by
PDFIG = xs
xs+xµviqr
QDFIG = −xµvidrxs+xµ
− v2
xµ
. (10)
4358
THANPISIT and NGAMROO/Turk J Elec Eng & Comp Sci
The generator motion equation is expressed by
ωm = (τm − τe)/2Hm
τe = ψdsiqs − ψqsids, (11)
where ωm is the state variable of rotor speed, τm and τe are respectively mechanical and electrical torques,
and Hm is rotor inertia.
Thus, the electrical torque τe can be expressed by
τe = xµ(iqrids − idriqs). (12)
The mechanical power pω extracted from the wind, which is a function of the wind speed vω , the rotor speed
ωm , and the pitch angle θp , can be approximated by
pω =ngρ
2Sncp(λ, θp)Arv
3ω, (13)
where ng is the number of machines in the wind park, ρ is the air density, Sn is the power rating, cp is the
performance coefficient or power coefficient, λ is the tip speed ratio, and Ar is the area swept by the rotor.
The cp(λ , θp) curve is approximated by
cp = 0.22(116
λi− 0.4θp − 5)e−
12.5λt , (14)
With1
λi=
1
λ+ 0.08θp− 0.035
θ3p + 1. (15)
The converter is modeled by an ideal current source, where iqr and idr are state variables and used for the rotor
speed control and the voltage control, respectively, as depicted in Figures 3 and 4. Differential and algebraic
equations for the converter currents are given by
Figure 3. Rotor speed controller.
iqr = (−xs+xµ
xµvp∗ω(ωm)/ωm − iqr)
1Te
idr = KV (v − vref )− v/xµ − idr
0 = v0ref − vref + vs
, (16)
where v0ref is the initial reference voltage, vs is an additional signal of the POD, p∗ω(ωm) is the power-speed
characteristic that roughly optimizes the wind energy capture, Kv is voltage control gain, Te is power control
4359
THANPISIT and NGAMROO/Turk J Elec Eng & Comp Sci
Figure 4. Voltage controller.
time constant, and imindr , imax
dr , iminqr , and imax
qr are minimum and maximum rotor currents in the direct axis and
the quadrature axis, respectively. Here, the stabilization of power oscillation is conducted by the voltage control
of idr with the POD signal. Without the risk of torsional oscillation in the wind turbine, the reactive power
output is used to suppress power oscillations.
The pitch angle control is illustrated in Figure 5 and described by
m
ref
p
0
1
p
p
K
T s
Figure 5. Pitch angle controller.
θp = (Kpϕ(ωm − ωref )− θp)/Tp, (17)
where ϕ is a function that allows varying the pitch angle set point only when ωm − ωref exceeds a predefined
value ±∆ω , ωref is the reference speed, Kp is the pitch control gain, and Tp is the pitch control time constant.
Based on Eqs. (11), (16), and (17), the fourth-order model of the DFIG can be represented by state
variables idr , iqr , ωm , and θp .
2.3. PSS and POD model
Figure 6 delineates the model of the PSS and POD as the second-order lead-lag compensator where KSTAB
is the gain, TW is the washout time constant and set at 10 s, and T1 , T2 , T3 , and T4 are time constants of
phase compensators. The input signals vSI of the PSS and POD are the speed deviation of the synchronous
generator and the local power flow signal, respectively. The PSS and POD output signals vS are limited by
±0.2 p.u. The PSS and POD output signals are sent to the AVR and the voltage-control loop of the DFIG,
respectively.
3. Proposed optimization technique
Conventionally, the parameters of the PSS and POD are optimally tuned at an operating point to achieve the
satisfied damping performance. Nevertheless, the resulting PSS and POD cannot confirm the stabilizing effect
4360
THANPISIT and NGAMROO/Turk J Elec Eng & Comp Sci
Gain Washout Lead-lag 1 Lead-lag 2
Figure 6. PSS and POD models.
over a wide range of operating points. To ensure the damping performance for diverse operating points, the
new parameter optimization for the PSS and POD by the probability method is described as follows.
The optimization objective is to obtain the damping ratio of all oscillation modes as more than 5% for
all possible operating points. Here, 1000 operating points, which are randomly generated by varying the power
outputs of all synchronous generators and load power between ±20% from a normal operating point, wind
speeds of the DFIG between 9 and 16 m/s, and a line outage contingency, are performed by MCS. The percent
damping ratio of the mth oscillation mode at the evth operating point (ζm,ev) can be calculated by
ζm,ev =−σm,ev√
σ2m,ev + ω2
m,ev
× 100; ev = 1,2,...,1,000, (18)
where σm,ev and ωm,ev are real and imaginary parts of the eigenvalue corresponding to the mth oscillation
mode at the evth operating point, respectively. Under 1000 events, the probability of the events for which the
percent damping ratio of the mth oscillation occurs between ζm and ζm+1% or Pr[ζm ,ζm+1%] is calculated
by
Pr [ζm, ζm + 1%] =N [ζm, ζm + 1%]
NA, (19)
where N [ζm ,ζm+1% ] is the number of events for which the percent damping ratio of the mth oscillation mode
occurs between ζm and ζm + 1%, and NA is the number of total events.
When ζmin and ζmax are the minimum and maximum percent damping ratios of each oscillation mode
for all events, respectively, the histogram of the occurring probability of each percent damping ratio for the
individual oscillation mode from ζmin to ζmax with 1% increasing step can be depicted as in Figure 7. To
make it easy for calculating the area under the histogram, a frequency polygon is established. In this study, the
acceptable percent damping ratio, which is set at 5%, is used to separate the histogram into two areas. Area1m
is the probability area where the percent damping ratio of the mth oscillation mode for all random 1000 events
is less than 5%. On the contrary, Area2m is the probability area where the percent damping ratio of the mth
oscillation mode for all random 1000 events is greater than 5%. The larger Area2m is, the higher damping
ratios of nine oscillation modes can be obtained. By this concept, the objective function for tuning parameters
of the coordinated PSS and POD is formulated by the minimization of the difference between Area1m and
Area2m for nine oscillation modes by
Maximize9∑
m=1
(Area2m −Area1m), (20)
4361
THANPISIT and NGAMROO/Turk J Elec Eng & Comp Sci
Area1m Area2m
Histogram
Frequency Polygon
AcceptableDamping ratio
yt ilib
ab
orP
Damping ratio (%). . . 1 2 3 4 5 6 . . .
0.1
0.2
0.3
Figure 7. Histogram of probability of % damping ratio.
Subjectto 0.1 < KSTAB < 50.00.1 < T1,2,3,4 < 1.0
.
PSO [17] is applied to solve optimal parameters in Eq. (20). The optimized PSS and POD are referred to as
PPSS+PPOD. The flowchart of the proposed optimization method can be delineated as in Figure 8.
The damping performance of PPSS+PPOD is compared with that of the PSS coordinated with the POD,
which is tuned at the normal operating point by
Minimize9∑
m=1
|ζspec − ζm|, (21)
Subjectto 0.1 < KSTAB < 50.00.1 < T1,2,3,4 < 1.0
,
where ζspec is the specified damping ratio, which is set at 5%. The PSS and POD optimized by Eq. (21) are
referred to as CPSS+CPOD.
4. Simulation results
The Power System Analysis (PSAT) toolbox [14] is applied for time-domain simulation. In the PSO, the number
of particles and iterations are set at 100 and 50, respectively.
The simulation study is divided into two parts. The first part compares the damping effects among PPSS,
PPOD, and PPSS+PPOD. Note that PPSS and PPOD are individually optimized by Eq. (20). The second
part compares the damping effect between PPSS+PPOD and CPSS+CPOD.
Figure 9 depicts the convergence curves of the objective function of Eq. (20) for PPSS, PPOD, and
PPSS+PPOD. Table 3 shows optimized parameters of CPSS+CPOD while Table 4 shows optimized parameters
of PPSS, PPOD, and PPSS+PPOD.
4.1. Comparison among PPSS, PPOD, and PPSS+PPOD
Figure 10 shows the probability of % damping ratios under 1000 operating points. The damping ratios in the case
of either PPOD or PPSS are higher than those of the system without PPOD and PPSS. Besides, PPSS+PPOD
yields the best damping performance for nine oscillation modes. PPSS+PPOD provides the probability of %
damping ratios that are greater than 5% for all oscillation modes. Figure 11 depicts the eigenvalues plot of all
oscillation modes under 1000 operating points. The result in Figure 11 satisfies that in Figure 10.
4362
THANPISIT and NGAMROO/Turk J Elec Eng & Comp Sci
Start End
Iter = 1
Generate the initial particle
Linearize power system
Compute damping ratio of
each oscillation mode by (18)
Evaluate the probability of the
events that the damping ratio of
individual oscillation mode
occurs in each bound by (19)
Use result from (19) to plot
histogram then convert to
frequency polygon.
Calculate objective function
(20) from frequency polygon
Get the optimal parameters
Iter >= Itermax
Itermax = 50
If gold > gnew then gbest = gold.
If gold < gnew then gbest = gnew
Select the maximum value of
objective function as gnew
Iter = Iter+1No
Yes
Update location and velocity in 100 particles
i = number of particles (set at 100)
xi = location for each particle
vi = velocity for each particle
α,β = acceleration constants
ε1,ε2 = random value between 0 to 1
gnew = the best value of objective function at iteration Iter
gold = the old best value of objective function
gbest = the maximum value from comparison between gnew and goldIter = iteration counter
Itermax = number of iterations
Figure 8. Flowchart of the proposed optimization method.
0 5 10 15 20 25 30 35 40 45 50-10
-9
-6
-4
-2
0
2
4
6
8
10
Iteration
no itcn uf e vi tcejbO
PPODPPSSPPSS+PPOD
Figure 9. Convergence curve of the objective function of Eq. (20).
A time-domain simulation is conducted to verify the damping effect of PPSS+PPOD. At the normal
operating point, it is assumed that the temporary three-phase fault occurs at bus 10 at t = 1 s for 100 ms and
is cleared naturally. Figure 12 shows simulation results of power flows between buses 1–2 and buses 9–39. The
4363
THANPISIT and NGAMROO/Turk J Elec Eng & Comp Sci
0 2 4 5 6 8 10 12 14 16 18 200
0.10.2
0 2 4 5 6 8 10 12 14 16 18 200
0.10.2
0 2 4 5 6 8 10 12 14 16 18 200
0.10.2
0 2 4 5 6 8 10 12 14 16 18 200
0.10.2
0 2 4 5 6 8 10 12 14 16 18 200
0.20.4
y tiliba
bor
P
0 2 4 5 6 8 10 12 14 16 18 200
0.10.2
0 2 4 5 6 8 10 12 14 16 18 200
0.10.2
0 2 4 5 6 8 10 12 14 16 18 200
0.20.4
0 2 4 5 6 8 10 12 14 16 18 200
0.5
Damping ratio (%)
Without controller PPOD PPSS PPSS+PPOD
MODE 1
MODE 2
MODE 3
MODE 4
MODE 5
MODE 6
MODE 7
MODE 8
MODE 9
Figure 10. Probability of % damping ratios under 1000 events.
Figure 11. Locus plot of all oscillation modes with 1000 events.
0 2 4 6 8 10 12 14 16 18 20-5
0
5Power !ow between buses 1-2
0 2 4 6 8 10 12 14 16 18 20-4
-2
0
2
4
Time (s)Pow
er !
ow (
p.u
.)P
ower
!ow
(p
.u.)
Power !ow between buses 9-39
Without controller PPOD PPSS PPSS+PPOD
Figure 12. Simulation results.
4364
THANPISIT and NGAMROO/Turk J Elec Eng & Comp Sci
Table 3. Parameters of CPSS+CPOD.
Parameters of CPSS+CPODKSTAB T1 T2 T3 T4
CPSS of G2 2.94 0.28 0.06 0.44 0.08CPSS of G8 48.66 0.22 0.09 0.91 0.09CPSS of G9 13.81 0.35 0.02 0.32 0.06CPOD of W1 9.22 0.21 0.04 0.21 0.04CPOD of W2 4.43 0.46 0.04 0.21 0.01CPOD of W3 7.15 0.51 0.03 0.37 0.03
Table 4. Parameters of PPSS, PPOD, and PPSS+PPOD.
Parameters of PPSSKSTAB T1 T2 T3 T4
PPSS of G2 38.35 0.79 0.06 0.31 0.04PPSS of G8 3.56 0.84 0.09 0.76 0.01PPSS of G9 8.27 0.24 0.02 0.45 0.05Parameters of PPOD
KSTAB T1 T2 T3 T4PPOD of W1 9.22 0.21 0.04 0.19 0.04PPOD of W2 4.43 0.46 0.04 0.21 0.01PPOD of W3 7.15 0.54 0.03 0.37 0.03Parameters of PPSS+PPOD
KSTAB T1 T2 T3 T4PPSS of G2 26.63 0.67 0.08 0.75 0.03PPSS of G8 31.37 0.54 0.08 0.35 0.09PPSS of G9 27.28 0.82 0.07 0.61 0.08PPOD of W1 10.62 0.43 0.03 0.75 0.01PPOD of W2 21.18 0.13 0.04 0.35 0.04PPOD of W3 5.76 0.14 0.04 0.24 0.03
undamped power oscillation occurs in the case without the controller. On the other hand, the power oscillation
can be suppressed by PPSS, PPOD, and PPSS+PPOD. Nevertheless, PPSS+PPOD provides the best damping
performance. These results confirm that the stabilizing effect of PPSS+PPOD is better than the individual
effect of PPSS or PPOD.
4.2. Comparison between CPSS+CPOD and PPSS+PPOD
Figure 13 shows the probability of % damping ratios of each oscillation mode for 1000 operating points.
PPSS+PPOD provides the damping ratios that are greater than the acceptable value (or ≥5%). On the
contrary, some damping ratios in the case of CPSS+CPOD are less than 5%. Moreover, the damping ratios in
the case of PPSS+PPOD are all higher than those of CPSS+CPOD. To confirm the result in Figure 13, the
scattering of eigenvalues for all oscillation modes under 1000 events is depicted in Figure 14.
Nonlinear simulations for case studies in Table 5 are carried out under various system uncertainties and
faults.
Figure 15 depicts power flows between buses 1–2 and buses 9–39 of each case study. In case 1, as
delineated in Figure 15a, severe power oscillation occurs in the case without the controller. On the other hand,
the power oscillation is completely damped by CPSS+CPOD and PPSS+PPOD. Figure 15b shows case 2,
4365
THANPISIT and NGAMROO/Turk J Elec Eng & Comp Sci
0 2 4 5 6 8 10 12 14 16 18 200
0.10.2
0 2 4 5 6 8 10 12 14 16 18 200
0.10.2
0 2 4 5 6 8 10 12 14 16 18 200
0.10.2
0 2 4 5 6 8 10 12 14 16 18 200
0.20.4
0 2 4 5 6 8 10 12 14 16 18 200
0.20.4
yti liba
bor
P
0 2 4 5 6 8 10 12 14 16 18 200
0.10.2
0 2 4 5 6 8 10 12 14 16 18 200
0.10.2
0 2 4 5 6 8 10 12 14 16 18 200
0.20.4
0 2 4 5 6 8 10 12 14 16 18 200
0.5
Damping ratio (%)
Without controller CPSS+CPOD PPSS+PPOD
MODE 1
MODE 2
MODE 3
MODE 4
MODE 5
MODE 6
MODE 7
MODE 8
MODE 9
Figure 13. Probability of % damping ratios under 1000 events.
-2.5 -2 -1.5 -1 -0.5 0
-10
-5
0
5
10
Real part (1/s)
Without controllerCPSS+CPODPPSS+PPOD
Damping ratio = 5%
Imag
inar
y p
art
(rad
/s)
Figure 14. Locus plot of all oscillation modes under 1000 events.
where the system is unstable in the case without a controller. PPSS+PPOD gives a better damping effect than
CPSS+CPOD. In cases 3 and 4, as shown in Figures 15c and 15d, CPSS+CPOD cannot stabilize the power
oscillations. On the contrary, PPSS+PPOD can suppress the power oscillation entirely. These simulation results
guarantee that the damping performance of PPSS+PPOD is superior to that of CPSS+CPOD.
5. Conclusion
This paper deals with the new optimization technique of a PSS and DFIG installed with a POD by the
probability method for stabilization of low-frequency electromechanical oscillations against multiple operating
points. MCS is applied to produce random operating events in the parameter optimization of the PSS and
4366
THANPISIT and NGAMROO/Turk J Elec Eng & Comp Sci
Table 5. Case study.
Case
Description
Applied disturbance
% Change
in power
output of all
generators
% Change
in power
output of
all loads
Wind
speed
of
DFIG1
Wind
speed
of
DFIG2
Wind
speed
of
DFIG3
N-1
line outage
contingency
1
Fault occurs at bus 39
for 150 ms and is
cleared naturally
+3% +2% 14 m/s 15 m/s 11 m/s Line 7–8
outage
2
Fault occurs at bus 7
for 150 ms and is
cleared naturally
No change +7% 12 m/s 14.5
m/s 9 m/s
Line 25 –26
outage
3
Fault occurs at bus 5
for 150 ms and is
cleared naturally
–5% No change 11 m/s 11 m/s 13 m/s Line 26 –27
outage
4
Fault occurs at line
16–21 at t = 1 s and is
cleared by tripping this
line at t = 1.10 s
without reclosing
No change +5% 14.5 m/s 10 m/s 15 m/s Line 14 –15
outage
(a) Case 1
(b) Case 2
(c) Case3
0 2 4 6 8 10 12 14 16 18 20-4
-2
0
2Power !ow between buses 1-2
0 2 4 6 8 10 12 14 16 18 20-4
-2
0
2
Time (s)
Power !ow between buses 9-39
Without controller CPSS+CPOD PPSS+PPOD
0 2 4 6 8 10 12 14 16 18 20-10
-5
0
5
10Power !ow between buses 1-2
0 2 4 6 8 10 12 14 16 18 20-10
-5
0
5
10
Time (s)
Power !ow between buses 9-39
Without controller CPSS+CPOD PPSS+PPOD
0 2 4 6 8 10 12 14 16 18 20-10
-5
0
5
10Power !ow between buses 1-2
0 2 4 6 8 10 12 14 16 18 20-10
-5
0
5
10
Time (s)
Power !ow between buses 9-39
Without controller CPSS+CPOD PPSS+PPOD
(d) Case 4
0 2 4 6 8 10 12 14 16 18 20-10
-5
0
5
10Power !ow between buses 1-2
0 2 4 6 8 10 12 14 16 18 20-10
-5
0
5
10
Time (s)
Po
wer
!o
w (
p.u
.)P
ow
er !
ow
(p
.u.)
Po
wer
!o
w (
p.u
.)P
ow
er !
ow
(p
.u.)
Po
wer
!o
w (
p.u
.)P
ow
er !
ow
(p
.u.)
Po
wer
!o
w (
p.u
.)P
ow
er !
ow
(p
.u.)
Power !ow between buses 9-39
Without controller CPSS+CPOD PPSS+PPOD
Figure 15. Simulation results of case studies 1 to 4: a) case 1, b) case 2, c) case 3, d) case 4.
POD under several generating and loading conditions, random wind speeds, and line outages. The optimization
is conducted by maximizing the probability that the desired damping ratios of all oscillation modes can be
achieved for all random operating points. Solving the optimization problem by PSO, the optimal parameters
4367
THANPISIT and NGAMROO/Turk J Elec Eng & Comp Sci
of the PSS and POD can be obtained automatically. Small-signal and transient stability results guarantee that
the damping performance of the proposed PSS and POD is much higher than that of the conventional PSS and
POD under multiple operating conditions and faults.
Acknowledgment
This work was supported by the King Mongkut’s Institute of Technology Ladkrabang Research Fund, No.
KREF115901.
References
[1] Prabha K. Power System Stability and Control. New York, NY, USA: McGraw-Hill Professional; 1994.
[2] Office of Energy Efficiency and Renewable Energy. Advantages and Challenges of Wind Energy. Washington, DC,
USA: EERE, 2017.
[3] Global Wind Energy Council. Global Wind Energy Outlook 2015. Brussels, Belgium: Wind Energy Council, 2015.
[4] Quintero J, Vittal V, Heydt G, Zhang H. The impact of increased penetration of converter-based generator on
power system modes of oscillation. IEEE T Power Syst 2014; 29: 2248-2256.
[5] Tazil M, Kumar V, Bansal RC, Kong S, Dong ZY, Freitas W, Mathur HD. Three-phase doubly fed induction
generators: an overview. IET Electr Power Appl 2010; 4: 75-89.
[6] Domınguez-Garcıa JL, Gomis-Bellmunt O, Bianchi FD, Sumper A. Power oscillation damping supported by wind
power: a review. Renew Sustain Energ Rev 2012; 16: 4994-5006.
[7] Fan L, Yin H, Miao Z. On active/reactive power modulation of DFIG-based wind generation for inter-area oscillation
damping. IEEE T Energ Convers 2011; 26: 513-521.
[8] Derafshian M, Amjady N. Optimal design of power system stabilizer for power systems including doubly fed
induction generator wind turbines. Energy 2015; 84: 1-14.
[9] Surinkaew T, Ngamroo I. Robust power oscillation damper design for DFIG-based wind turbine based on a specified
structure mixed H2/H∞ control. Renew Energ 2014; 66: 15-24.
[10] Surinkaew T, Ngamroo I. Power system oscillations damping by robust decentralized DFIG wind turbines. J Electr
Eng Technol 2015; 10: 487-495.
[11] Huang H, Chung CY. Coordinated damping control design for DFIG-based wind generation considering power
output variation. IEEE T Power Syst 2012; 27: 1916-1925.
[12] Surinkaew T, Ngamroo I. Coordinated robust control of DFIG wind turbine and PSS for stabilization of power
oscillations considering system uncertainties. IEEE T Sustain Energ 2014; 5: 823-833.
[13] Ackermann T. Wind Power in Power System. Chichester, UK: Wiley & Sons, 2005.
[14] Milano F. Power System Analysis Toolbox Version 2.1.8. Dublin, Ireland: University College Dublin, 2013.
[15] Zhou EZ, Malik OP, Hope GS. Theory and method for selection of power system stabilizer location. IEEE T Energ
Convers 1991; 6: 170-176.
[16] Rahim AHMA, Habiballah IO. DFIG rotor voltage control for system dynamic performance enhancement. Electr
Power Syst Res 2011; 81: 503-509.
[17] Kennedy J, Eberhart R. Particle swarm optimization. In: Proceedings of the IEEE Neural Network Conference; 27
November–1 December 1995; Perth, Australia. New York, NY, USA: IEEE. pp. 1942-1948.
[18] Vladislav A. Induction Generators for Wind Power. Brentwood, UK: Multi-Science Publishing Co., 2005.
4368
THANPISIT and NGAMROO/Turk J Elec Eng & Comp Sci
Appendix
DFIG data [18]: blade length: 75 m; blade number: 3; air density: 1.225 kg/m3 ; inertia time constant = 3 s;
gearbox ratio = 1/89; stator resistance = 0.01 p.u.; rotor resistance = 0.01 p.u.; stator reactance = 0.1 p.u.;
rotor reactance = 0.08 p.u.; magnetizing reactance = 3 p.u.; number of poles = 4; voltage control gain = 50;
power control time constant = 0.01; pitch control gain = 50; pitch control time constant = 3.
1
top related