Sunteți pe pagina 1din 6

Coordinated Tuning of Synchronous Generator

Controllers for Power Oscillation Damping


Amer Mešanović Ulrich Münz Rolf Findeisen
Otto von Guericke University of Magdeburg Siemens AG Otto von Guericke University
Siemens AG, Munich, DE Princeton, USA Magdeburg, DE
amer.mesanovic@siemens.com ulrich.muenz@siemens.com rolf.findeisen@ovgu.de

Abstract—The increasing share of renewable generation will Another important property of renewable generation is their
lead to time-varying oscillatory modes in large scale power volatility, leading to time varying oscillation modes which
systems. Today, oscillatory modes are defined by the location change depending on the percentage of renewable generation
of conventional power plants and by the current load level in
the grid. In the future, more conventional power plants will be and its geographic location in the grid. [2], [7]. In order to
periodically disconnected depending on the share and location of damp time varying oscillations, the PSSs of generators would
renewable generation. Hence, the oscillatory modes will become have to be tuned conservatively, leading to a degraded system
more time variable than today, e.g. depending on the weather. performance [8].
Thus, new functions become necessary in order to optimally use Time varying oscillatory modes are the main motivation
existing controllers, e.g. power system stabilizers, for damping
of inter-area oscillations. In this paper, we present a method for for the contribution of this work: we propose a method for
parameter tuning of generator controllers for optimal oscillation coordinated tuning of generator controllers with structured
damping based on structured H-infinity optimization. We assume H∞ optimization and depending on the current grid state.
that power system stabilizers are not appropriately parametrized The tuning of the PSSs is coordinated because local and
and use the proposed method to find the optimal parametrization uncoordinated design of controllers may degrade oscillation
of the controllers. The efficiency of the proposed method is shown
with detailed rounded-mean-square simulations of a well known damping in the system [9]. We assume that the system is
four generator and the IEEE 39 bus power system. already in a state in which power oscillations are not suffi-
Index Terms—Distributed control, H∞ optimization, power ciently dampened. The tracking and the update of controller
systems, robustness parameters for changing operating conditions is beyond the
scope of this work. For this purpose, a power system model
I. I NTRODUCTION with detailed controller and generator models and with explicit
Power system control today is built around the properties dependencies on the controller parameters is first defined. Then
and dynamics of synchronous generators: dispatchable control the model is linearized around the steady-state and the H∞
of their generation and large time constants due to their inertia. norm of the system is optimized by using the bounded-real
To further improve the small signal stability of the system, i.e. lemma and PK iteration for the parameter tuning [10].
the ability of a power system to remain in synchronism after In Section II, the complete dynamic model of the power
being subjected to a small disturbance, some generators were system is presented. The optimization method is described in
equipped with a power system stabilizer (PSS). The oscillation detail in Section III, and the effectiveness of the proposed
modes today shift due to the periodical load fluctuations, method is shown in Section IV. In the end, the conclusion
and the PSSs are manually tuned to damp these modes. and future steps are given in Section V.
For example, in Europe there are four dominant inter-area A. Literature review
oscillation modes today [1].
Various works consider power oscillation damping in power
The percentage of renewables in the power system is
systems. The authors in [11]–[13] consider H∞ optimization,
constantly increasing,leading to significant changes in power
H2 optimization, or pole placement for the controller synthe-
system dynamics [2]. Distributed generators, as opposed to
sis. Sensitivity analysis is used in [5], [14]–[16], whereas [17],
conventional generators, have either very small, or no inertia
[18] apply LQR controllers for oscillation damping. Various
and damping properties, which makes them react much faster
other methodologies can be found in [8], [19]–[22] etc. How-
to contingencies, however most do not react at all. As the share
ever, most of works either consider simplified power system
of distributed generation in the grid increases, the impact of
models [3]–[5], [14], [17], [19], [20], or new controllers are
low inertia and damping effects increases [3]–[6].
added on top of the existing power system model [7], [8],
The authors acknowledge funding from the German Federal Ministry for [11]–[13], [18] which makes practical application complex
Economic Affairs and Energy within the funded research project DynaGrid- and expensive. Very few works consider the optimization of
Center. existing controller parameters in the system [15]. So far, no
978-1-5386-1953-7/17/$31.00
c 2017 IEEE structured optimization approaches, except iterative simulation
A. Synchronous generator model
To model the synchronous machine dynamics, a 6th order
model is used, as described in detail in [31], given with
δ̇ = ω − ωs (1a)
2H Pm
ω̇ = − Ψd iq + Ψq id − Dω (1b)
ωs ω
Fig. 1. A synchronous generator together with the modelled controllers. Ψ̇f d = −Ef d + Rf d if d Ψ̇1d = −R1d i1d (1c)
Ψ̇1q = −R1q i1q Ψ̇2q = −R2q i2q (1d)
methods, for the detailed structured controller synthesis of vd = −Ψq ωs − Ra id vq = Ψd ωs + Ra iq (1e)
power system controllers have been presented. ψd = −Ld id + Lmd (if d + i1d ) (1f)
Structured controller synthesis, considering a fixed con- ψq = −Lq iq + Lmq i1q + Lmq i2q (1g)
troller structure, has been a topic of significant research ψf d = Lf d if d + Lmd (−id + i1d ) (1h)
in the past decades and a variety of approaches has been
developed. Approaches include non-smooth optimization tech- ψ1d = L1d i1d + Lmd (−id + if d ) (1i)
niques [23], [24], sum-of-squares techniques [15], [25] etc. ψ1q = L1q i1q − Lmq iq + Lmq i2q (1j)
Very popular approaches are iterative linear matrix inequality ψ2q = L2q i2q − Lmq iq + Lmq i1q , (1k)
(LMI) approaches [26]–[28]. In this paper, we use the so-
called iterative coordinate descent method or P-K-iteration where Pm is the mechanical power from the turbine, Ψd and
[29]. Previous work of the authors [30] considers H∞ , H2 and Ψq are the direct and quadrature components of the main
pole placement for power oscillation damping with renewable flux, Ψf d is the excitation circuit flux, Ψ1d and Ψ1q are
generation. However, very simplified generator models were the d- and q-axis amortisseur 1 fluxes, Ψ2q is the q-axis
used, and in this work we use detailed generator and con- amortisseur 2 flux, R, L and i with appropriate sub-indices
troller models, leading to significantly larger problems. We are the resistance, inductance and current, respectively, in the
furthermore compare our model with nonlinear simulations in corresponding circuit, vd and vq are the d- and q-component
a commercial power system simulation software. of the terminal voltage, Ef d is the excitation voltage, ω is
the rotational frequency of the generator rotor, Ra is the
amortisseur resistance, Lmd and Lmq are the d- and q-axis
B. Contributions main inductances, and δ is the rotor angle relative to the
coordinate system rotating with the nominal angular frequency.
In this work, we first present the analytical derivation of
a detailed power system model with an explicit dependency B. Controller models
on the tunable parameters. Most of the work presented in the
The governor controls the turbine’s mechanical power out-
literature review works either with simplified power system
put Pm into the synchronous machine based on the angular
models [3]–[5], [14], [17], [19], [20], or the power system
velocity of the generator. The two-state steam turbine and
model is used as a “black box” model to which new controllers
governor model TGOV1 is chosen, which represents the
are added [7], [8], [11]–[13], [18]. Furthermore, we propose
motion of steam through the reheater and turbine stages. It
coordinated tuning of existing PSS parameters for power os-
is a standard model used in stability studies [11], [32], and
cillation damping by using structured H∞ optimization. This
the model is presented in Fig. 2.
approach optimally uses existing controllers and no additional
The generator exciter controls the terminal voltage of the
controllers are required. Finally, we present a time-domain
generator via the field winding voltage Ef d . We use here
comparison of the linearized model to the detailed nonlinear
the one-state excitation system illustrated in Fig. 3 [11],
model of the power system, showing good accuracy of the
[14], however more complex exciters can be used as well.
linearization and the optimization results in the detailed model.
The inputs to the exciter are the reference voltage Vref , the
generator terminal voltage VT , and the input from the power
II. DYNAMIC POWER SYSTEM MODEL system stabilizer VP SS . The output of the exciter is the field
winding voltage Ef d . The exciters objective is a constant
In this section, the dynamic power system model for the op- terminal voltage of the generator, however it can reduce the
timization is defined. Figure 1 illustrates how the synchronous damping properties of a synchronous generator [31].
generator is coupled with its controllers. The generator is To increase the oscillation damping, some generators are
controlled via the mechanical power Pm from the turbine, equipped with an analogue or digital power system stabilizer
which is controlled by the governor, and the field voltage Ef d (PSS). Several measured values can be chosen for the input,
from the exciter. The dynamic models of the generator with its and here the deviation from the nominal angular frequency is
controllers, as well as the power flow equations, are presented chosen. Figure 4 shows the PSS model, see [11], [14], [31],
subsequently. and this paper assumes all of its parameters to be tunable.
Fig. 4. The dynamic model of the simple power system stabilizer from [11],
[14], [31], where KS is the PSS gain, TW is the washout time constant,
T1 -T4 are the lead-lag filters time constants. All of the PSS parameters are
optimized.

Fig. 2. The dynamic model of the steam turbine and governor TGOV1 [32]. where pi and qi are the injected active and reactive power into
The frequency droop gain of the governor KP is optimized. the i-th bus in the grid, vi and θi are the magnitude and angle
of the voltage phasor at the i-th bus, and G and B are the
conductance and susceptance matrix of the grid [31].
E. Coupled dynamic power system model
The generator models, consisting of the synchronous gen-
erator, governor and turbine, exciter, and a PSS, presented in
Section II, are coupled with the power flow equations (2) and
Fig. 3. The dynamic model of the exciter [11], [14]. The exciter gain KA is a differential-algebraic nonlinear model is obtained
optimized
ẋ =f (x, w, K) 0 = h(x, w, K), (3a)
were x is the vector of all generator states in the model,
The controllers are described with a total of six states and w is the vector of external inputs into the system, e.g. the
the synchronous generator also has six states. Therefore, each apparent power of the system loads, K is the vector of tunable
synchronous generator with its controllers is modeled with a parameters of all generators, defined in Subsection II-C, f
total of twelve states. represents the generator dynamics, and h represents the power
flow equations. Because we are interested in the small-signal
C. Tunable Parameters
behavior of (3), we linearize the system around the steady-
We define eight tunable parameters for each synchronous state value x0 obtained by solving f (x0 , w0 , K) = 0 and
generator, marked with red in figs. 2 to 4: h(x0 , 0, w0 , K) = 0, and we obtain a linear system
• the proportional gain KP of the steam turbine governor,
ẋ = A(K)x + B(K)w (4)
which determines how aggressive the governor reacts to
frequency deviations. The accuracy of the linearization is shown on a numerical
• the proportional gain KA of the exciter, which determines example subsequently.
how aggressive the exciter reacts to voltage deviations. III. O PTIMIZATION METHOD
• all parameters of the PSS: KS , TW , T1 , T2 , T3 , T4 .
The H∞ -norm of a system represents the maximal am-
The parameter KP is optimized to show the possibilities of plification of amplitude of any harmonic input signal in any
the proposed method and because it has a significant influence output direction [28]. In case of a single-input single-output
on the system stability [33]. This parameter today is used system, the H∞ norm is the maximum of the system Bode
for primary-reserve allocation without consideration for the magnitude plot. Thus, by minimizing the H∞ -norm of a
system dynamics, however this could change in the future. system, we automatically improve the damping of the least
Other parameters are assumed to be fixed. damped oscillatory mode of the power system. This in turn
minimizes the oscillations of the defined outputs after, e.g., a
D. Power Flow Equations load step.
Interarea and interplant oscillations are considered, which As the H∞ norm characterizes the input-output behavior of
usually have a frequency of below 2 Hz [31]. Thus, the the system, appropriate performance inputs and outputs need
dynamics of the power lines, whose time constants are much to be defined. Since our goal is the minimization of active
smaller, can be neglected. For this reason, the power system power oscillations in the system, which manifest through the
with N buses and M power lines is modeled using algebraic oscillation of generator frequencies, we define the generator
power flow equations angular frequencies as the system performance outputs
T
N
X  y = ω1 . . . ωnG = C(K)x, (5)
pi = |vi ||vj | Gij cos(θi − θj ) + Bij sin(θi − θj ) (2a)
j=1
where nG is the number of generators in the power system.
N
The active power of the load in several buses is defined as the
performance input, i.e.
X 
qi = |vi ||vj | Gij sin(θi − θj ) − Bij cos(θi − θj ) , (2b)
T
j=1 w = P1 . . . PnL , (6)
TABLE I
I NITIAL PARAMETRIZATION OF THE TUNABLE CONTROLLER PARAMETERS
FOR EACH GENERATOR .

Gov Exc PSS


Gen KP KA Ks TW T1 T2 T3 T4
G1 10 10 10 20 0.019 0.04 8.1 10.8
G2 20 16 16 12.5 0.012 0.025 12.96 6.75
Fig. 5. A two-area system from [31, p. 813, Example 12.6]. G3 50 22 22 9.1 0.009 0.018 17.82 4.91
G4 30 28 28 7.1 0.007 0.014 22.68 3.86

where nL is the number of loads in the power system.


TABLE II
The system G(K) = (A(K), B(K), C(K)) is now defined O PTIMAL H∞ NORM PARAMETRIZATION OF THE TUNABLE CONTROLLER
and we minimize the H∞ norm of the system from w to y PARAMETERS FOR EACH GENERATOR .

by tuning K. The tuning is done with help of the bounded Gov Exc PSS
real lemma [10], and the following optimization problem is Gen KP KA Ks TW T1 T2 T3 T4
obtained G1 137 7.94 6.67 4.05 0.017 0.041 7.84 3.39
G2 147 5.78 6.65 3.99 0.032 0.025 12.4 4.87
min γ (7a) G3 147 24.22 26.59 2.47 0.015 0.018 18.4 3.26
P,K G4 142 21.04 29.06 1.81 0.007 0.014 22.87 4.19
A(K)T P + P A(K) P B(K) C(K)T
 

s.t.  B(K)T P −γI D(K)T  ≺ 0


C(K) D(K) −γI parameters are presented in Table I. We consider a load step
(7b) of 100 MW in bus 9 in Fig. 5 as the disturbance input w for
Kmin ≤ K ≤ Kmax P = P T  0, (7c) the simulation. All generators are equipped with PSSs in this
example because we consider them as equivalent models for
where the notation ‘ 0’ and ‘≺ 0’ for a matrix means several generators.
the matrix is positive definite or negative definite, Kmin and Figures 6 and 7 show the linear (dashed lines) and non-
Kmax are the minimal and maximal bounds for the tunable linear (solid lines) simulations of generator frequencies and
parameters, respectively, and ≤ / ≥ is defined element-wise terminal voltages with the initial parameters from Table I.
for vectors. The obtained optimization problem is non-convex, The nonlinear simulation is done in Simscape Power Systems
however, it can be solved efficiently with state-of-the-art with detailed nonlinear models of the synchronous generators
solvers [34] by iteratively optimizing P and K, thus the name and the nonlinear power flow. As shown in Fig. 6 and 7,
PK iteration. It shows very good results for the analyzed power the difference in responses is small and the linear model
systems, as evident from the subsequent numerical examples. can be used for the optimization of the detailed nonlinear
Another advantage of this method is that it can be readily model. The linear system consists of 48 states and 32 tunable
adapted to communication failures. For example, if a generator parameters described in Section II-C. Figures 8 and 9 show
loses communication to the central unit, the generator can keep the system frequency and voltage magnitude response after the
the last parameters it obtained, or revert to predefined default parameter optimization. The tuned parameters are shown in
parameters. Then, in the subsequent optimization, we can set Table II. A much better system response is obtained by tuning
the controller parameters of the respective generator to this the controller parameters. The simulation with the optimized
defined value, and optimize the parameters of other generators. parameters of the linear model shows good correspondence
In this way, we still achieve ”optimal“ POD under the given to the detailed nonlinear simulation. Tuning the controller
circumstances, even in the case when the communication to parameters reduced the system’s H∞ norm by a factor of 30.
several generators is lost. The pole-zero map of the system with the initial and opti-
IV. N UMERICAL EVALUATION
In this section, we show the effectiveness of the proposed
G1 linear
method on the four generator system from [31] and the IEEE G2 linear
frequency (Hz)

ten generator 39 bus system. 60 G3 linear


G4 linear
G1 nonlinear
A. Four generator system G2 nonlinear
59.95 G3 nonlinear
First we consider the grid from [31, p. 813, Example 12.6] G4 nonlinear
presented in Fig. 5. The parameters of the transmission grid,
5 10 15 20 25
as well as the parameters of synchronous generators, can be
time(s)
found in [31]. The parameters of the TGOV1 controller are
T1 = 0.04s, T2 = 1s, T3 = 2s, and Dt = 0, chosen to satisfy Fig. 6. Generator frequency response after a 100 MW load step in bus 9 in
Fig. 5 with initial parameters presented in Table I. The solid lines represent
the limits defined in [35], while the exciter time constant the simulation with the linear model, whereas the dashed lines represent the
TA is set to 1s as in [31]. The initial values for the tunable simulation with the nonlinear model.
G1 linear G1 linear
G2 linear G2 linear
20 G3 linear 20 G3 linear

VT (kV)
VT (kV)

G4 linear G4 linear
G1 nonlinear G1 nonlinear
19.5 G2 nonlinear 19.5 G2 nonlinear
G3 nonlinear G3 nonlinear
G4 nonlinear G4 nonlinear
19 19
5 10 15 20 25 5 10 15 20 25
time(s) time(s)
Fig. 7. Generator voltage magnitude response after a 100 MW load step Fig. 9. Generator voltage magnitude response after a 100 MW load step
in bus 9 in Fig. 5 with initial parameters presented in Table I. The solid in bus 9 in Fig. 5 with tuned parameters presented in Table II. The solid
lines represent the simulation with the linear model, whereas the dashed lines lines represent the simulation with the linear model, whereas the dashed lines
represent the simulation with the nonlinear model. represent the simulation with the nonlinear model. The y-axis scale is set to
be the same as in Fig. 7.

G1 linear
G2 linear
frequency (Hz)

60 G3 linear
G4 linear
G1 nonlinear
G2 nonlinear
59.95 G3 nonlinear
G4 nonlinear

5 10 15 20 25
time(s)
Fig. 8. Generator frequency response after a load step in bus 9 in Fig. 5 with
tuned parameters presented in Table II. The solid lines represent the simulation
with the linear model, whereas the dashed lines represent the simulation with
the nonlinear model. The y-axis scale is set to be the same as in Fig. 6.

mized parameters is presented in Fig. 10. With H∞ controller


tuning, the system poles were shifted to the left to achieve
better power oscillation damping. Note that the power system
does not have a swing bus. Therefore, the phase angles of all Fig. 10. Pole-zero map of the system which shows the system poles and
generators continuously increase, even in steady-state. These zeros with initial parameters (blue) and optimized parameters (red).
dynamics are represented by the single pole at the origin,
which does not have an influence on the inter-area oscillations.
The efficiency of the method was shown on a four generator
B. Ten generator IEEE 39 bus system and IEEE 39 bus power system, in which power oscillations
We also consider the 10 generator IEEE 39 bus system, were effectively dampened with parameter optimization and no
shown in Fig. 11, to evaluate the method for larger systems. additional controllers. As the proposed method can be time-
The linear model of the system consists of 120 states and 80 consuming to solve for very large systems, as future work we
tunable controller parameters. We define a 100 MW load step will consider faster methods for the optimization including
in bus 2 in Fig. 11 as the performance input for the simulation. model reduction [15], we will evaluate the method on larger
Figures 12 and 13 show the simulation results with the
linear model for the inital and optimized parameters. Because
the accuracy of the linear model was verified in the previous
example, the simulation with the nonlinear model is not shown
in Fig. 12 and 13. Again, the system H∞ norm was reduced
by a factor of 40 just by tuning the controller parameters.

V. C ONCLUSIONS
In this paper, a structured H∞ algorithm is proposed which
minimizes the oscillations of generator frequencies after a load
change by only tuning existing controllers. For this purpose,
a detailed analytic linear model of the system with explicit
parameter dependencies is defined and its accuracy is verified
with nonlinear simulation. The method can be readily extended
to a wide variety of controller structures and generator types. Fig. 11. The New England IEEE 39 bus ten generator system.
60 Gen 1
Gen 2 [12] X. Wu, F. Dörfler, and M. Jovanović, “Input-output analysis and de-
frequency (Hz) Gen 3 centralized optimal control of inter-area oscillations in power systems,”
59.99 Gen 4 IEEE Trans. Power Systems, vol. 31, no. 3, pp. 2434 – 2444, 2016.
Gen 5
Gen 6 [13] S. Schuler, U. Münz, and F. Allgöwer, “Decentralized state feedback
59.98 Gen 7 control for interconnected systems with application to power systems,”
Gen 8 Journal of Process Control, vol. 24, no. 2, pp. 379–388, 2014.
Gen 9 [14] M. Mahmoudi and K. Tomsovic, “A distributed control design method-
59.97 Gen10
ology for damping critical modes in power systems,” Proc. IEEE Power
5 10 15 20 25 and Energy Conference at Illinois (PECI), 2016.
[15] B. Marinescu, B. Mallem, H. Bourles, and L. Rouco, “Robust coor-
time (s) dinated tuning of parameters of standard power system stabilizers for
Fig. 12. Generator frequency response after a 100 MW load step in bus 2 in local and global grid objectives,” in PowerTech, 2009 IEEE Bucharest.
Fig. 11 with initial controller parameters. IEEE, 2009, pp. 1–7.
[16] L. Rouco, “Coordinated design of multiple controllers for damping
60 Gen 1 power system oscillations,” International Journal of Electrical Power
Gen 2 & Energy Systems, vol. 23, no. 7, pp. 517–530, 2001.
frequency (Hz)

Gen 3 [17] M. Mahmoudi, J. Dong, and K. Tomsovic, “Application of distributed


59.99 Gen 4 control to mitigate disturbance propagations in large power networks,”
Gen 5
Gen 6 North American Power Symposium (NAPS), 2015.
59.98 Gen 7 [18] R. Preece, J. Milanovic, A. A. M., and O. Marjanovic, “Damping of
Gen 8 inter-area oscillations in mixed AC/DC networks using WAMS based
Gen 9 supplementary controller,” IEEE Trans. Power Systems, vol. 28, no. 2,
59.97 Gen10
pp. 1160 – 1169, 2013.
5 10 15 20 25 [19] T. Graber, S. Hoehn, and M. Luther, “A contribution to thorough com-
prehension of POD provided by FACTS devices,” Proc. International
time (s) ETG Congress, 2015.
Fig. 13. Generator frequency response after a 100 MW load step in bus 2 [20] K. Liao, Z. He, Y. Xu, G. Chen, Z. Dong, and K. Wong, “A sliding
in Fig. 11 with optimized controller parameters. The y-axis scale is set to be mode based damping control of DFIG for interarea power oscillations,”
the same as in Fig. 12. IEEE Trans. Sustainable Energy, vol. 8, no. 1, pp. 258 – 267, 2017.
[21] F. J. De Marco, N. Martins, and J. Ferraz, “An automatic method
for power system stabilizers phase compensation design,” IEEE Trans.
Power Systems, vol. 28, pp. 997–1007, 2013.
systems with other component types, and the tracking and [22] A. Yaghooti, M. Buygi, and M. Shanechi, “Designing coordinated power
the update of the controller parameters for changing operating system stabilizers: A reference model based controller design,” IEEE
Trans. Power Systems, vol. 31, no. 4, pp. 2914 – 2924, 2016.
conditions will be considered as well. [23] S. Gumussoy, D. Henrion, M. Millstone, and M. L. Overton, “Multiob-
jective robust control with HIFOO 2.0,” Proc. 6th IFAC Symposium on
R EFERENCES Robust Control Design, vol. 42, no. 6, pp. 144–149, 2009.
[24] P. Gahinet and P. Apkarian, “Decentralized and fixed-structure H∞
[1] E. Grebe, J. Kabouris, S. Lopez Barba, W. Sattinger, and W. Winter, control in MATLAB,” IEEE Conference on Decision and Control and
“Low frequency oscillations in the interconnected system of continental European Control Conference, pp. 8205 – 8210, 2011.
europe,” IEEE PES General Meeting, 2010. [25] C. Hol and C. Scherer, “A sum-of-squares approach to fixed-order H∞
[2] S. Al Ali, T. Haase, I. Nassar, and H. Weber, “Impact of increasing synthesis,” Lecture Notes in Control and Information Science, 312.
wind power generation on the north-south inter-area oscillation mode in Springer, 2005.
the European ENTSO-E system,” IFAC Proceedings Volumes, vol. 47, [26] J. Han and R. Skelton, “An LMI optimization approach for structured
no. 3, pp. 7653–7658, 2014. linear controllers,” 42nd IEEE International Conference on Decision and
[3] A. Ulbig, T. S. Borsche, and G. Andersson, “Impact of low rotational Control, no. 5, pp. 5143 – 5148, 2004.
inertia on power system stability and operation,” IFAC World Congress, [27] A. Karimi, H. Khatibi, and R. Longchamp, “Robust control of polytopic
vol. 19, no. 1, pp. 7290 – 7297, 2014. systems by convex optimization,” European Cont. Conf. (ECC), 2007.
[4] B. Poolla, S. Bolognani, and F. Dörfler, “Placing rotational inertia in [28] C. W. Scherer, “Structured H∞ optimal control for nested intercon-
power grids,” American Contr. Conf. (ACC), pp. 2314 – 2320, 2016. nections: A state-space solution,” Systems & Control Letters, pp. 1105–
[5] T. Borsche, T. Liu, and D. J. Hill, “Effects of rotational inertia on power 1113, 2013.
system damping and frequency transients,” in Decision and Control [29] E. Simon, V. Wertz, P. Rodriguez-Ayerbe, C. Stoica, and D. Dumur,
(CDC), 2015 IEEE 54th Annual Conference on. IEEE, 2015, pp. 5940– “LMIs-based coordinate descent method for solving BMIs in control
5946. design,” IFAC Proc. Volumes, vol. 18, no. 1, pp. 10 180–10 186, 2011.
[6] M. Garmroodi, D. Hill, G. Verbic, and J. Ma, “Impact of tie-line power [30] A. Mešanović, U. Münz, and C. Heyde, “Comparison of H∞ , H2 , and
on inter-area modes with increased penetration of wind power,” IEEE pole optimization for power system oscillation damping with remote
Trans. Power Systems, vol. 31, no. 4, pp. 3051 – 3059, 2016. renewable generation,” IFAC Workshop on Control of Transmission and
[7] R. Preece and J. Milanovic, “Tuning of a damping controller for multi- Distribution Smart Grids CTDSG, vol. 49, no. 27, pp. 103–108, 2016.
terminal VSC-HVDC grids using the probabilistic collocation method,” [31] P. Kundur, Power System Stability and Control. McGraw-Hill, 1993.
IEEE Trans. Power Delivery, vol. 29, no. 1, pp. 318 – 326, 2014. [32] P. Pourbeik, “Dynamic models for turbine-governors in power system
[8] A. Fuchs, M. Imhof, T. Demiray, and M. Morari, “Stabilization of large studies,” Technical report PES-TR1, 2013.
power systems using VSC-HVDC and model predictive control,” IEEE [33] I.-Y. Chung, W. Liu, D. Cartes, E. Collins, and S. Moon, “Control
Trans. Power Delivery, vol. 29, no. 1, pp. 480 – 488, 2014. methods of inverter-interfaced distributed generators in a microgrid
[9] M. Gibbard, D. Vowles, and P. Pourbeik, “Interactions between, and system,” IEEE Trans. Industry Applications, vol. 46, no. 3, pp. 1078
effectiveness of, power system stabilizers and facts device stabilizers in – 1088, 2010.
multimachine systems,” IEEE Trans. Power Systems, vol. 15, no. 2, pp. [34] J. Lofberg, “YALMIP : a toolbox for modeling and optimization in MAT-
748 – 755, 2000. LAB,” IEEE International Conference on Robotics and Automation, pp.
[10] C. Scherer and S. Weiland. (2017, Feb) Linear matrix inequalities 284 – 289, 2004.
in control, lecture notes. TU Delft. [Online]. Available: http: [35] NEPLAN AG. (2016) Turbine-governor models. [On-
//www.dcsc.tudelft.nl/∼cscherer/lmi/notes05.pdf line]. Available: http://www.neplan.ch/wp-content/uploads/2015/08/
[11] M. Raoufat, K. Tomsovic, and S. Djouadi, “Virtual actuators for wide- Nep TURBINES GOV.pdf
area damping control of power systems,” IEEE Trans. Power Systems,
vol. 31, no. 6, pp. 4703 – 4711, 2016.

S-ar putea să vă placă și