EMTP-MATLAB/ SPS Co-Simulation for Large

advertisement
EMTP-MATLAB/ SPS Co-Simulation for Large-Scale Integration of Wind Power Plants
C.F. Mugombozi1, O. Saad1, T. Roudier3, V. Morissette2, R. Gagnon1 and G. Sybille1
1 : Institut de Recherche d'Hydro-Québec (IREQ)
2 : Hydro-Québec TransÉnergie
3 : Kiastek
Correspondence:
Email: mugombozi.chuma-francis@ireq.ca
Institut de Recherche d’Hydro-Québec (IREQ)
1800 blvd. Lionel Boulet
Varennes (Québec), Canada, J3X 1S1
This paper presents an approach for reducing simulation time of large electric systems with detailed
dynamic modelling of control systems. Simulation of transients such as faults takes hours when detailed
models of wind generator (WG) or High Voltage Direct Current (HVDC) link have to be considered.
Detailed models are needed to adequately analyze dynamics of wind generator integrated to electric
systems [1]. Moreover, one may require replication of models from a particular simulator to another
simulator. This requires both resources and knowledge across diverse simulation softwares, thus increasing
costs and work time for set up and deployment of simulation tools.
Two aspects of this approach tackle issues presented above: 1) Parallel simulation and 2) Hybrid
simulation. Parallel simulation is first introduced. Electric power network is decoupled on multiple
subnetworks assigned to a particular simulator. All simulators run in parallel on multi-core computer.
Variables are exchanged via a co-simulation bus. While some simulators incorporate parallelization and
multithreading concepts, this approach extends multi-core processor advantage to any simulator. Second,
hybrid simulation is introduced. Different simulators are combined and participate to a single simulation of
large electric systems. Usually interoperability and portability of simulators require careful handling of data
representation equivalency using standard language such as Common Information Model (CIM) or using
Dynamic Link Library (DLL). Standardization of model data representation has evolved but is not yet
widely used. Also, while many simulators incorporate such DLL capability [2], its implementation is not a
trivial task for average user. The approach presented in this paper provides usage of preferred (native)
model in any simulation software but integrates it with other simulators.
Demonstration of parallel and hybrid simulation is made by combining EMTP-RV [3] and MATLAB/SPS
[4] simulators via a CosiMate bus [5]. The main power network is modelled as a single instance either in
EMTP or MATLAB/SPS while up to 72 aggregated wind power plants (WPP) are modelled in several
instances either in EMTP or MATLAB/SPS. Hybrid combinations are also made with instances of different
simulators.
Challenges faced for decoupling network and combining various simulators include model and solver
equivalencies, data transfer and synchronization, and stability and convergence of decoupled large electric
systems. Following steps were carried out to achieve our goal.
Decoupling is achieved using controlled voltage and current sources. This technique was previously
applied in the Hypersim real time simulator [6] developed by Hydro-Québec’s research institute (IREQ).
However this decoupling technique introduces a time step delay between subnetworks. When particular
models such as transmission lines use a natural delay, they can be used as decoupling elements. In our
approach we assume no sufficient natural time delay. This is consistent with latest cases of integration of
wind power plant (WPP) connected to a bus [1]. The collector network is mainly modelled in one simulator
as lumped capacitors, inductors and filters thus allowing suitable connection to current sources from the
external simulator modelling the wind power plant. In addition a prediction/correction mechanism is
introduced to compensate time step delay across simulators due to measure of voltage and current
transformed into controlled sources. We start by applying this technique on a simple power network
available in MATLAB/SPS demos [4] and replicated for both simulators. The wind generator model used is
the average model from [1].
Models and solver equivalencies of EMTP and MATLAB/SPS simulators are achieved by setting
similar method of solution, namely trapezoidal integration method, and by modifying a number of control
parameters (proportional and integral gains) or changing control strategies, for instance, for reactive energy
in WPP model, to minimize as best as possible initialization discrepancies between EMTP and
MATLAB/SPS solvers. Other sources of discrepancies between simulators include modelling of Phaselocked loop (PLL) elements, measuring blocks and simultaneity of solution of control and power network
equations. Impacts of possible differences of simulators on synchronicity of data exchanged trough the
interface have to be considered. Synchronization is assessed by comparing instantaneous power injected by
the wind generator from MATLAB/SPS to EMTP, and vice versa. Time step is set to 50 or 25 microseconds for all cases. All single-simulator and hybrid cases match very well for all signals including
instantaneous power and DC voltage inside wind generator model.
Several tests were performed on a 21-bus simplified model of Hydro-Québec (HQ) system (Fig. 1) and
on modified IEEE 39-bus systems with detailed model of generators. In order to analyze impact of different
computation loads on CPU, 6, 12, 32 and 72 WPPs were respectively connected to HQ 21-bus and IEEE
39-bus while maintaining initial generated power on buses. Each WPP is divided in two within the same
EMTP instance or MATLAB instance. For dynamic analysis a ten cycles fault is applied near bus 33 at 10
seconds for 20 seconds time frame and near MTL bus at 0.2 seconds for 4 seconds time frame for
respectively modified IEEE 39-bus and HQ 21-bus systems. The reference simulation times are cases
without decoupling.
As expected, best performances are obtained when computation load is equally shared between processors.
Also, gain in simulation time diminishes when any core is over-loaded thus becoming a bottle neck for data
exchange in whole system. All Gains obtained by proposed approach are presented in Fig.2. A maximum
gain of 7 is obtained for case 14 running on 8 core CPU computer. Still some improvement for generic
decoupling method is needed. Nonetheless gain obtained may reduce work time required for simulation and
by ricochet contribute to reduction of overall cost involved.
References
[1] R. Gagnon, G. Turmel, C. Larose, J. Brochu, G. Sybille, M. Fecteau, Large-Scale Real-Time
Simulation of Wind Power Plants into Hydro-Québec Power System, 9th International Workshop on
Large-Scale Integration of Wind Power into Power Systems as well as on Transmission Networks for
Offshore Wind Power Plants, 18-19 October 2010, Quebec city, Quebec, Canada
[2] O. Saad, Computation of power system transients: modeling portability using EMTP-RV DLL, Panel
session on interoperability and portability, IEEE Power and Energy Society General meeting, 2011,
Detroit
[3] DCG-EMTP, EMTP-RV 2.2, Electromagnetics Transients Program: http://www.emtp.com.
[4] MATLAB/Simulink, The MathWorks Inc, Software Package for Modeling and Analyzing Dynamic
Systems: http://www.mathworks.com
[5] CosiMate, Kiastek: http://www.cosimate.com
[6] V-Q. Do, J-C. Soumagne, G. Sybille, G. Turmel, P. Giroux, G. Cloutier, S. Poulin, Hypersim, an
integrated real-time simulator for power networks and control systems, Third International Conference
on Digital Power System Simulators - ICDS 99, Vasteras, Sweden, 1999
QUE7a
QUE7b
QUE7c
cC
MIC7c
MIC7b
C
Cs SAG
bB
Cs
SAG7
B
LNOE
aA
bB
cC
aA
PSS_select
Reseau
Nord-Est
A
CHM7c
CHM7b
CHM7a
SIMPLIFIED
HQ 21-BUS
Reseau
Nord-Ouest
A
B
C
1
MTLc
STATUT STABILISATEURS:
PSS=0 hors service;
PSS=1 Puissance (dPa);
PSS=2 Multibande (dw)
MTLb
MTLa
powergui
MIC7a
Discrete,
Ts = 5e-005 s.
LNEO
SAG
PSS
Cs
CHM-JCA
Cs CHMS
QUE7
aA
bB
cC
c
b
a
C
B
c
b
Xfo3
C
a
B
A
C
B
Xfo2
A
MTL-QUE
A
c
b
a
MTL7
Xfo1
INTERFACE WPP
Three-Phase Fault
A
Eoliennes Mtl
m
m
A
B
B
Vf _
C
Eoliennes Qc
C
C
A
Vf
B
A
A
B
C
Q-MTL
A
F-QUE
C
B
A
INTERFACE WPP
A
B
C
A
B
C
B_5 MTL
Regulateurs5
C
Pm
B
Pm
B
F-MTL
C-MTL
5%Pnom
C
Eoliennes Mtl1
Eoliennes Qc1
A
B
C
======================================================================
c
A
c
C
c
C
A5
A
B25
575V
-
25 kV/ 575 V
1.75MVA
3
exin_disc
v2.0
b
s
c
em
FROM OTHER SIMULATOR
B5
+
C
Iabc_25_Qc
Vabc_25_Qc
-
B
DATA TRANFER PORT:
CURRENT INPUTS
aB
s
b
+
a
B
DATA TRANFER PORT:
VOLTAGE OUTPUTS
C
A
2
exout_disc
v2.0
1/z
A
B
INTERFACE
WPP
&
POWER NETWORK
1
Filter1
Q=50
======================================================================
[Vdc3]
<Vdc_V>
[wr3]
<wr_pu (IG speed)>
12
Wind (m/s)
m
0
[Vabc_25_QC]
[Q_pu3]
<Q_pu>
Qref _pu
Vabc (V)
[P_pu3]
<P_pu>
[Iabc_25Qcl]
Iabc (A)
DFIG Wind Turbine6
MATLAB SIMULINK BLOCK OF WPP
EMTP-RV BLOCKS OF WPP
AND PREDICTION CORRECTION
Figure 1 Simplified system of HQ 21-bus interfaced via CosiMate ports with MATLAB/SPS and EMTP Model of WPP
8 Core CPU
20
16
11
7
5
3
21
9
17
2
1
case no
case no
4 Core CPU
0
1
2
3
4
Gain = T standard / T CosiMate
14
22
15
18
13
23
8
10
4
6
19
0
2
4
6
Gain = T standard / T CosiMate
8
Figure 2 Gains in simulation times (T standard = not decoupled)
Download