IEEE Industry Applications Society Annual Meeting New Orleans, Louisiana, October 5-9, 1997
FACTS MODELLING AND CONTROL : APPLICA'TIONS TO THE INSERTION OF A STATCOM ON POWER !SYSTEM P. PETITCLAIR, Y. BESANGER, S. BACHA, N. HADJSAID Laboratoire d'Electrotechnique de Grenoble LEG - INPG(ENSIEG) / UJF - CNRS UMR 5529, ENSIEG - LEG, BP 46, F-38402 Saint Martin d'Heres Cedex, France Phone : (33) 4-76-82-62-85, Fax : (33) 4-76-82-63-00, E-mail :
[email protected] WWW : http//www-leg.ensieg.inpg.fr/ ABSTRACT - Much attention has been focused recently on FACTS (Flexible AC Transmission System) devices. Among these devices, the STATCOM (STA Tic Compensator) contributes to the voltage support especially in terms of static and dynamic stability. This paper proposes three developed models with control structure for the STATCOM. At first, a comparative study for dynamic operations is presented. Afterwards, the performances of these models, when the STATCOM is inserted in a power system, are studied. For the models, the CPU time requirements have been compared, 1. INTRODUCTION
Much attention has been focused recently on FACTS devices (Flexible AC Transmission System)(l] to improve the speed control of line parameters (voltage, impedance and phase). Among these devices, shunt compensation using Power Electronics allows power systems to improve their performances. Shunt compensation contributes to the voltage support especially in terms of static [2] and dynamic stability. Several projects are currently planned [3] or are close to prototype installation [4] concerning a particular shunt FACTS device which is called "STATCOM" (STATic CO M pe nsato r) . The STATCOM is a non linear variable structure device. Therefore, a robust control of this structure has to be designed in a nonlinear way. There are different structures of STATCOM [5] involving different dynamics. This paper proposes three developed models with control structure for the STATCOM. At first, a comparative study for dynamic operations is presented, which includes a basic, a topological and a generalized averaged models. In this part, the study has been carried out with the MATRlXx program. Afterwards, the performances of these models, when the STATCOM is inserted in a power system, are studied by using the EUROSTAG program [6]. This software is dedicated to the simulation of power system dynamics. For this purpose, three drastic situations are examinated : line outages, loss of generators and short-circuits. For the models, the CPU time requirements have been compared. Results have
been obtained on the 30 bus New England network [7]. 2. Dynamic Modelliing of the STATCOM
A. Basic Model The basic model consists on representing the static characteristic of the STATCOM (Line voltage E versus reactive current I) (figure 1). The reactive power compensator is considerecl as perfect (without losses and dynamics). In this case, the dynamics are performed only by the control law, due to the fact that the STATCOM modelling do not take into account the whole structure of the device. This model is widely used for power system studies.
I
it-
1-i;
ICmax ILmax I :STATCOM Static Characteristic
Figure
E
L
A
Figure 2 :Basic Model of the STATCOM
0-7803-4067-1/97/$10.000 1997 IEEE. 2213
0 power system
complex zeros of the system are different depending on the operating point [8].
B. Topological Model of the inverter This model is used to describe the dynamic behaviour of the STATCOM, with its power electronics structure (Figure 3). At first, only one inverter structure of the STATCOM is studied. Equations of the system are dependent on the variable structure of the inverter. The three phases current and voltage are function of the switch states.
t (SI Figure.4 :/q versus time for a step of a Figure. 3 : Topo/ogica/scheme of the converter
In order to separate the active and reactive values, the
Park transformation has been used. Iq represents the reactive current and the continuous voltage Vdc. The active value, particularly the active current Id, represents the system losses. This discontinous and nonlinear model is necessary to understand the behaviour of the STATCOM with its structure. However, this model can not be used to build a classical control law, but allows continuous models to be built.
A robust non linear control law is designed in order to control the behaviour of the reactive current Iq (2).For example a linearization via feedback [8,9]control allows a perfect control of the reactive current to be obtained. The speed response is given by h . An external voltage loop is necessary to obtain a static characteristic of the STATCOM. For this loop a PI linear controler is tuned taking into account the power system dynamics.
C.Generalized Averaged Model (GAM) The topological model is used to obtain the generalized averaged model [8].The averaged one is time invariant but nonlinear (1).
X=
w
X+-L LS
0 1
]
E
(1)
U
The STATCOM basic model can be used to study the behaviour of the Power system, in the case of the perfect STATCOM without losses and dynamics. The topological model describes perfectly the STATCOM dynamics as a function of the switch states. However it can not be used in a power system network simulation software. On the other hand, the generalized averaged model is well adapted for the simulation of the dynamic behaviour of the STATCOM on power systems. This model describes the STATCOM with the inverter dynamics. The effect of the inverter losses can also be studied with this model, which can not be done with the currently used basic model.
with X = 3. Insertion of the STATCOM Models in EUROSTAG
In fact, it is dependent on the firing angle a of the switch with reference to the first phase voltage zero crossing. It also allows the dynamic behaviour of the topological model to be well conserved. Figure 2 gives Iq behaviours for Topological and Generalized Averaged Models. A study of this kind of system has showed that a linear controller, such as a classical PI (Proportionnal and Integral controller) can not be robust, since the
A. GAM validation A dynamic behaviour of the generalized averaged
model has been validated with the topological model using MATRlXx program [ l o ] in section 2. The generalized averaged model is implemented in a software dedicated to the simulation of power system
2214
dynamics : EUROSTAG [ 6 ] . A simple test network allows the STATCOM performance to be validated. The same condition with the averaged model are simulated with the EUROSTAG and the MATRlXx programs to validate the implemented STATCOM model. An ideal generator is connected to the STATCOM via an impedance line. An active power load is connected to the STATCOM (figure 5). The STATCOM used is based on an 80 MVar [ l l ] .
The STATCOM generalized averaged model is described as in with its control figure 7. The controls are tuned to have a time response of the current Iq equal to 100ms with a damping factor of 0.7. Then, the proportional integral corrector of the voltage loop keeps the values Kp=20 for the proportional gain and Ki=40 for the integral gain. The proportional gain of the current loop keeps the value h=20. Then the responses of both generalized averaged STATCOM models in MATRIXx and in EUROSTAG are compared for a step of load. Figures (7) and (8) show the results obtained in this case of the voltage E and reactive current Ig respectively. Voltages and currents are given in per-unit. Curves are identical for simulations carried out with Matrixx and EUROSTAG which validate the implementation of the STATCOM model in EUROSTAG.
Bus Connection
Line
t
1
Figure 5 :test conditions
I
y'-g
' , if I 1
bw 1
LineazionH via feedback Invei-ter
m Id
Io
w e 2
System
I
I I I I I
I I I I I
I
I
external Volt:age Loop
I
I 1
PU
l'ool
1
0.99
0.98
0.97
0.96
0.95 10.0
10.1
0.2
0.25
0.3
0.35
t (SI1
10,z
Figure 7-a :Voltage E with EUROSTAG
Figure 7-b : Vollage E with MATRlXx
221 5
0.4
3.45
I
.4-
0.4
(------'I:
i / ;
....,..,. . ........... .................................,,.,.,,,,.,.,,,,.....,...............,...................................,,,,,,,...,,,..,
.3-
3.25 0.2
3.15
- r . . . . , . . . . , . . . . , .j. . . 1
0.25
0.2 10 0
10.1
Figure 8-a :Current lq with EUROSTAG
0.35
0.3 t (SI
10 2
0.4
Figure 8-b : Current 1q with MATRlXx
B. Power System cases
C.Short-Circuit
Tests have been performed on the 30 buses NewEngland system (figure 9). Power System behaviour is analyzed for the basic and generalised averaged models. The basic model has the same synopsis but without the internal current loop and consists of a variable impedance (susceptance B, Figure 2). The voltage loop controler has the same values than the generalized averaged model.
Figure 10 shows the voltage at bus 8 in the case of a three-phase to ground solid short-circuit at bus 6. The fault has a duration of 100 ms.
:
PU 1.21
.8
.o
'
200.00
200.05
with Basic Model
I
with GAM
200.10
200.15
figure 10: Lbus 8 voltage
Figure 9 :New-EnglandNetwork
At first, a slowly graduated increase of load at bus 8 has showed that the static characteristic (voltage U versus STATCOM current) is the same for the two models. Three drastic situations are then examinated : line outages, loss of generators and short-circuits.
When the short-circuit is cleared, the basic model reacts dangerously while the voltage at bus 8 reaches the value of 1.12 p.u. This could result in a voltage unstability during the simulation where as, in fact, the problem does not occure in practice. At the same time, the GAM produces small voltage oscillations which not affect the network stability : the generalized averaged model takes into account the converter dynamics and active losses. The GAM also leads to large power oscillations on lines (Figure 11) when the basic model is not able to take into account this phenomena.
221 6
PU
with Basic Model
F.Calculation time
I
The introduction of the inverter (with its active losses) and its non linear control in the STATCOM model results in a longer calculation time. Depending on the test nature and for the same tests durations, the total CPU time is about 1.3 up to 2.7 times longer for the generalized averaged model compared to the basic model.
4. CONCLUSION -4.
'
, 200.00
20d.05
200.10
200.15
200.20
In this paper, three different STATCOM models have been tested in a dyn,amic power system simulation program. The basic maldel is convenient for the steadystate stability studies whereas the averaged model allows the transient stability tests. Moreover, the averaged model takes in account active losses of the STATCOM and thus, 1,s more accurate. On the other hand, the introduction of this model results in a longer calculation time.
206.25
Figure 1 1 :Active power on Lbus 7-Lbus8line
D.Line Outages Figure 12 shows the result for a line outage. In this case, the voltage drop is more important for the generalized averaged model than for the basic model. The responses are presented for the two models, the generalized averaged model presents oscillations.
5. IReferences PU
,980
[ l ] N. G. HINGORANI : "FACTS - Flexible AC Transmission System" EPRl FACTS conference, Cincinnati, Ohio, Nov. 14-16,1990.
. Q78
[2] Y. BESANGER. J. C. PASSELERGUE. N. HADJSAID, R. FEUILLET : "lmprovment of power system performance by inserting FACTS devices" IEE ACDC'96, London, UK, 29 April-3 May, 1996.
with Basic Model
[3]R. J. KOESSLER, B. FARDANESH, M. I. HENDERSON, R. ADAPA : Feasibility studies for STATCON application in New York state" EPRl FACTS conference 3,Baltimore, Maryland, Oct. 5-7, 1994.
"
[4] C. SCHAUDER, T. W. CEASE, A. EDRIS and al. : "+/- 100 MVAR static condenser installation for TVA substation" EPRl FACTS conference 3, Baltimore, Maryland, Oct. 5-7, 1994.
,972
[5] C. HOCHGRAF, R. LASSETIER, D. DIVAN, T. A. L I P 0 : "Comparison of multilevel inverters for static var compensation" IEEE IAS, Oct. 1994, pp 921-928.
.Q S 8
161 EUROSTAG user's manual : Electricite De France - Direction des etudes et recherches, Departement MOS. TRACTEBEL, EnergyEngineering, Depattement Reseaux.
with G.A.Model .964
[7] M. A. PA1 '
" Energy function analysis for power system stability" Kluwer academic publishers, Boston, Mass., 1989
Rgure 12 :Lbus 8 Voltage
[8] P. PETITCLAIR. S. BACHA. J. P. ROGNON : "Averaged modelling and nonlinear control of an ASVC (Advanced Static Var Compensator)" IEEE PESC96, Baveno, Italy, June 24-27.1996.
E. Generator outage In the case of a loss of a generator, the voltage drop at bus 8 is identical for the two models but the effects of the inverter dynamics delay the voltage support for the GAM. In this case, the comparison of the two models leads to the same conclusions.
[9] A. ISlDORl : "Nonlinear control systems - an introduction" Springer Verlag, second edition, 1989.
[lo]Integrated System Inc. "MATRIXx product Family",january 1996 [I I ] Electric Power Research Institute : "Development of an Advanced static VAR compensator" Contract RP3023-1
2217
