Full PPR ISGT Dinamarca Final

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

Comprehensive approach for prediction and assessment

of power system transient stability in real-time


Diego E. Echeverría1, José L. Rueda2, Jaime C. Cepeda1, Delia G. Colomé3 and István Erlich2
1
Corporación CENACE, Quito, Ecuador
2
University Duisburg-Essen, Duisburg, Germany
3
Universidad Nacional de San Juan, San Juan, Argentina
{decheverria, jcepeda}@cenace.org.ec, {jose.rueda, istvan.erlich}@uni-due.de, colome@iee.unsj.edu.ar

Abstract-- Secure operation of power systems is receiving a clustering for identification of critical machines, and the
renewed interest due to considerable structural changes and emergency single machine equivalent method (E-SIME)
greater uncertainties. Thus, motivated by implementation of method for transient stability assessment [3]. Based on a set of
wide-area measurement systems in power systems worldwide, measurements acquired in real time after occurrence and
recent research work is being devoted to the development of clearing of a disturbance or any sudden outage (i.e. immediate
enhanced tools for fast security assessment within reduced time post-fault conditions), it provides relevant information
frames. The work presented in this paper introduces an regarding the number and alignment of critical machines, the
approach that exploits the post-disturbance data acquired via system stability margin, and the time to instability. Tests are
phasor measurement units (PMUs) by using a finite difference
performed on the New England benchmark power system.
based method for the prediction of rotor angles, data-mining-
based clustering for identification of critical machines, and II. MODEL OF MULTI-MACHINE POWER SYSTEM
single-machine-infinite bus (SMIB) transformation to evaluate
the system transient stability in terms of unstable margin and Mathematically, for each synchronous generator in a
time to instability. Numerical results obtained by applying the power system, the rotor angle i (i=1,2,...n) is determined by
approach on the New England test system demonstrates the the swing equation [4]:
feasibility and effectiveness that could be achieved in predicting d  i (t )
the critical machines and unstable conditions, which is also of  i ( t )   o
great value for defining suitable emergency control actions. dt (1)
d i (t ) 1
Index Terms—Equivalent machine, PMU, Transient stability,   Pmi (t )  Pei (t ) (2)
real-time assessment. dt Mi
where M is the moment of inertia, Pm is the mechanical
I. INTRODUCTION power input, Pe is the electrical power output, and ω is the
speed of the generator rotor.
There are many instances in which power systems are Equations (1) and (2) can be approximated by using the
operated close to their security limits. Moreover, under such
trapezoidal rule. Thus, the values of speed and rotor angles
operating conditions, some faults and contingencies may lead
to major consequences, including widespread disruptions or for each generator, at each time instant tk, are determined as
even blackouts, whose root causes have been occasionally follows [5]:
attributed to large-disturbance (i.e. transient) stability t  tk  tk-1 (3)
problems [1]. The emergence of synchrophasor and wide area
Pm i (tk )  Pei (tk )
acquisition technologies has become crucial for the i (tk )  t (4)
development of innovative technological solutions regarding Mi
improved real-time monitoring, and control tasks, which aim i (tk )  i (tk-1 )  i (tk ) (5)
at maximizing the operating flexibility while fulfilling
1
economic, security and environmental requirements. i (tk )  i (tk-1 )  i (tk-1 )  i (tk ) t (6)
Remarkably, tracking the system post-fault dynamic 2
performance in real time is enabled nowadays by phasor III. EQUIVALENT MACHINE METHOD
measurement unit devices (PMUs), which should be
strategically distributed throughout the system to achieve high Broadly speaking, the single machine equivalent method
observability [2]. (SIME) assesses the behavior of a power system in its post-
fault configuration (i.e. after a disturbance inception and its
This paper proposes an approach that combines a novel
algorithm for rotor angle prediction, data-mining-based clearance) in terms of a generalized single-machine-infinite
bus (SMIB) transformation [6]. This SMIB equivalent results constitutes the reactance between the machine terminal and
from the aggregation of “critical machines” and “non-critical the point in the network where the PMU device is installed
machines” into two equivalent machines. whereas Xq is the quadrature-axis reactance. The sampling
The identification of these two groups of machines is interval t = 20 ms is to be used in this paper (i.e. to match
illustrated in Fig. 1(a). The time-dependent parameters of the typical PMU sampling rate at 50 Hz).
“generalized SMIB” (i.e. rotor angle and speed, mechanical
and electrical powers), are computed from the parameters of
the power system machines. In E-SIME, these multi-machine
parameters are derived from real-time measurements, and
recalculated at a regular interval.

(a) (b)
Figure 1. SIME’s principle: (a) Power systems curves and identification of
critical machines; (b) Equal area criterion application to the SMIB equivalent
system.

The stability properties of the SMIB are inferred from the


application of the equal-area criterion (EAC). This yields two
essential pieces of information, namely, stability margin and
critical machines. The stability margin indicates the
imbalance between accelerating and decelerating areas
belonging to SMIB’s P- plane, or, equivalently, of the Pa-
plane, as illustrated in Fig. 1(b), where Pa denotes the SMIB
accelerating power:
Pa  Pm  Pe (7)
Pm is mechanical power and Pe is electrical power.
IV. PROPOSED APPROACH
The overall procedure of the proposed approach is
schematically depicted in Fig. 2.
Figure 2. Flowchart of the proposed approach.
A. Calculation of rotor variables
The values of the machine angles  at each time instant tk B. Prediction of rotor angles
are calculated from measurements of magnitude and phase of This task is accomplished by using the finite difference
voltages and currents provided by PMU devices located at method to predict the rotor angle position at a later time point
system high voltage buses. This procedure is necessary since tp [9]. An expression for estimating the angle  at tp is
PMU devices do not provide information that can be directly obtained by applying approximations for the first and second
used for transient stability analysis [7]. Such devices measure derivative of the Taylor series expansion of the rotor angle
electrical variables in real time (e.g. voltages and currents), discrete equation truncated to the third order and provided
whereas mechanical variables have to be estimated for that values of  at the instants tk, tk-1 and tk-2 are known:
carrying out transient stability assessment (e.g. generator i (tp )  i (tk-2 )  1 (tp  tk-2 )   2 (tp  tk-1 )(tp  tk-2 )
rotor angle and speed). Hence, the proposed approach uses (9)
the methodology presented in [8] to approximate the where
generator rotor angles. Basically, the equation to be used is:  (t )  i (tk-1 )  (t )  i (tk )  
1  i k-2 , 0  i k-1 , 2  1 0 (10)
  X q +X e  Icos  tk-2  tk-1 tk-1  tk tk-2  tk
  tan -1   (8)
 Vt +  X q +X e  Isin  C. Determination of critical machines
 
Once the rotor angles are predicted, the machines are
where the terminal current I, the terminal voltage Vt, and the
ranked according to their angles and the candidate critical
voltage angle , are available signals from PMU devices. Xe machines are identified by considering the largest angular
distance between two successive values of rotor angle. The If the unstable conditions are not reached, the steps
suitable aggregation of these machines provides the candidate described in the previous subsections IV.B, IV.C and IV.D are
SMIB equivalent [3]. For this purpose, an angle  dissimilarity repeated using a new measurement set. Otherwise, the
matrix (D) is built by applying the Recursive Method for candidate SMIB is a critical one, for which the following
Online Coherency Calculation (RMOCC) [10]. items are computed successively:

0 d12 ... d1n   The critical angle u.


d 0 ... d 2 n   The margin to instability u, given by
D   21 (11) u
 . . ... .  1
  P d   2 M
2
  a k (17)
 d n1 dn2 ... 0  k

 The time to instability


where n is the number of buses, and dij is the distance between u
d
bus i and bus j determined via RMOCC [10]. tu  tk  
k 2

(18)
After building the dissimilarity matrix, the corresponding   Pa d  2k
M k
similitude matrix (Q) can be obtained by (12), where I is the
identity matrix and 1 is the all-ones vector [2]. These matrices where k and k are the rotor angle and speed at time tk.
represent the variability between elements.
F. Emergency control
1 1   1  The prediction of the time to instability is of great
Q   I  11 D I  11 (12)
2 n   n  importance to arrive at appropriate control decisions. As
shown in Fig. 2, the proposed procedure for prediction and
Using eigenvalues (Λr) and eigenvectors (Vr) of the assessment of transient stability can be coupled with a
matrix Q, it is possible to calculate their corresponding scheme for emergency control, for which, it is crucial to
principal coordinates [2], [11], and put them together to form
determine the type, location, extent, and execution time of
a multi-dimensional space data matrix (X), as shown by (13)
potential corrective actions. For illustrative purposes, this
and (14). Yr is a matrix that contains the new orthogonal
variables that keep the original distances (principal paper considers generation shedding as a countermeasure.
coordinates) [2], [11].
V. RESULTS
Yr  Vr Λ r1/ 2 (13)
The proposed approach is tested on the New England
system [12], whose single line diagram is shown in Fig. 3.
X  Yr (14)
The implementation was accomplished by using DIgSILENT
Once the data matrix X ( stands for rotor angle data) is Power Factory© [17] along with several routines written in
determined, its multi-dimensional data points are grouped MATLAB® [18]. The case study presented in this section
into two clusters using the Fuzzy C-Means (FCM) algorithm. focuses on the prediction of transient instability after the
This clustering algorithm allows partitioning the machines occurrence of a three phase short circuit in line 2-25 near bus
into critical and non-critical groups. 2. The prediction calculations start once a fault is cleared at
t=0.22 s and consider at least three measurements at the time
D. Inferring the Pa- curve of the SMIB equivalent
points tk-2t, tk-t, tk. Results for every stage of the
Based on the predicted angles of the individual machines, procedure are summarized in the following subsections.
the values of the candidate SMIB, specially, its accelerating
power Pa and rotor angle δ, are determined for at least three A. Computation and prediction of the rotor angles
successive data sets acquired at tk-2Δt, tk-Δt and tk [3]. Equation Fig. 4 shows the time evolution of rotor angles of
(15) is written for the three different time instants and solved system’s generators, which are derived, according to the
for a, b, and c. guidelines given in subsection IV.A, from measurements of
voltages and currents provided by PMUs (placed at high
Pa    a2  b  c (15)
voltage buses of each generator’s transformer). For a given
E. Prediction and assessment of system instability time point, the rotor angles are predicted for the next100 ms
Equation (15) is solved to determine whether the SMIB’s by using (9) and (10), cf. subsection IV.B. Fig. 5 shows the
angle reaches unstable conditions, i.e. Pa (tu )  0, Pa (tu )  0 , accuracy that can be achieved by comparing the predicted
where Pa is the SMIB accelerating power and tu is the time to values with the theoretical ones derived from PMU data. The
instability. average values were removed from the calculated angles
shown in Fig. 4 to obtain those shown in Fig. 5.
Pa ( u )  a u 2  b u  c  0 (16)
B. Identification of critical machines D. Evaluating the system transient stability
The procedure given in subsection IV.C is followed to At this stage, equation (16) is solved to compute the
build the dissimilarity matrix D, the corresponding similitude unstable angle u in order to ascertain whether an unstable
matrix Q, and the data matrix Xθ in order to partition the condition has been reached (i.e. critical SMIB). If unstable
machines into critical and non-critical groups based on FCM condition does not occur, the steps described in the previous
clustering. Fig. 5 also illustrates the grouping and relative subsections IV.B, IV.C and IV.D are continually repeated
resemblance between the time evolution of the rotor angles of using an updated and expanded measurement set. For
the generators belonging to each one of the two groups. illustrative purposes, Table I lists values for different
subsequent instants tk (points of time at which a measurement
is received) as well as the corresponding estimated unstable
margin  and time to instability tu. The results listed in the
Table were obtained after detecting an unstable condition,
which occurs after 540 ms in this particular case. This is
reasonable since the application of the procedure given in
subsections IV.E would yield η >0 (stable condition) for the
first few measurements. Note that the time tu, in which the
instability is predicted (i.e. first negative value of η), is
considerably short (around 620 ms). Thus, for tk=560 ms,
there would be approximately 60 ms remaining to perform a
control action, which suggest the implementation of a scheme
at the control center for execution of countermeasures in an
automatic manner.

Figure 3. New England test System.

Fig. 5. Prediction of the rotor angles from previously calculated values and
identification and grouping of critical and non-critical generators.

Fig. 4. Comparison between the rotor angle values derived from PMU
measurements and those obtained via model-based time domain simulations.
The signals correspond to post-disturbance responses after the occurrence of
a three phase short circuit in line 2-25 near bus 2. The fault is cleared at
t=0.22 s.
C. Pa- curve of the SMIB equivalent
After identifying the critical machines, the next step aims
at determining the parameters of the candidate SMIB, that is,
the accelerating power Pa and rotor angle δ at the instants tk-
2Δt, tk-Δt and tk. This task is repeated if the new data of a
subsequent point of time is added to the already collected
information. The resulting estimated Pa-δ curves are Fig. 6. Prediction of the Pa-δ curve of the SMIB equivalent
illustrated in Fig. 6. These curves present better
E. Performing emergency control
approximations with respect to the real curve (red dotted
continuous line) when more measurement points are The design of suitable control actions involves increasing
considered by the prediction algorithm. the deceleration area of the system, which could be achieved,
for instance, by using predefined countermeasure schemes,
such as generation shedding [15].
For this case study, emergency control is performed by conditions, which is also of great value for defining suitable
tripping generators G8 and G9 at t=610 ms (i.e. 50 ms after emergency control actions. For illustrative purposes, this
the instability is first detected), which constitute the critical work considered countermeasures in the form of generation
machines (cf. Fig. 5). Fig. 7 shows the evolution of the rotor shedding, which would not be desirable in real practice.
angles when considering the execution of the resulting Nevertheless, future research work is being directed towards
emergency control action, which is decided by the alert the development of a more sophisticated control scheme,
provided by the proposed approach. Note that the loss of which will be tested on a real power system. Further research
synchronism and the collapse of the system are avoided. work is also needed to cover practical implementation
TABLE I concerns related to communication speed and reliability,
Prediction and assessment of transient stability since the available time for application of the approach in real
time is very short and more sophisticated information
tk (s)  tu (s) technology development should be reached.
0.540 >0 -
REFERENCES
0.560 -1.161 0.620
[1] P. Kundur, J. Paserba, V. Ajjarapu, G. Andersson, A. Bose, C.
0.580 -2.323 0.627 Canizares, N. Hatziargyriou, D. Hill, A. Stankovic, C. Taylor, T. Van
0.600 -3.218 0.633 Cutsem, and V. Vittal, “Definition and classification of power system
0.620 -3.793 0.638 stability IEEE/CIGRE joint task force on stability terms and
definitions”, IEEE Transactions on Power Systems, Vol. 19, No. 3,
0.640 -4.015 0.642 Aug. 2004, pp. 1387 – 1401.
[2] J.C. Cepeda, J.L. Rueda, I. Erlich, and D.G. Colomé, "Probabilistic
Approach-based PMU placement for Real-time Power System
Vulnerability Assessment," in Proc. 2012 IEEE PES Innovative Smart
Grid Technologies (ISGT) Europe Conference, Berlin, Germany, Oct.
2012.
[3] M. Glavic, D. Ernst, D. Ruiz-Vega, L. Wehenkel, and M. Pavella, “E-
SIME- A Method for Transient Stability Closed-Loop Emergency
Control: Achievements and Prospects”, in Proc. IREP Symposium, Bulk
Power System Dynamics and Control VII, Charleston South Carolina,
USA, 2007.
[4] P. Kundur, “Power System Stability and Control”, New York:
McGraw-Hill, USA, 1994.
[5] F. Milano, “Power System Modelling and Scripting”, Berlin: Springer,
Germany, 2010.
[6] Y. Zhang, L. Wehenkel, P. Rousseaux and M. Pavella, “SIME: A
Hybrid Approach to Fast Transient Stability Assessment and
Contingency Selection”, EPES, Vol.19, No.3, 1997, pp. 195-208.
[7] A. Del Angel, P. Guerst, D. Ernst, M. Glavic, and L. Wehenkel,
“Estimation of Rotor Angles of Synchronous Machines Using Artificial
Fig. 7. Evolution of generator rotor angles when performing emergency Neural Networks and Local PMU-based Quantities”, Neurocomputing,
control in the form of generation shedding: tripping of generators 8 and 9 at Vol. 70, 2007, pp. 2668–2678.
610 ms. [8] E. Ghahremani, M. Karrari, M. B. Menhaj, O. P. Malik, “Rotor Angle
Estimation of Synchronous Generator from Online Measurement”
VI. CONCLUSIONS Universities Power Engineering Conference, 2008, UPEC 2008.
Generator speed and rotor angle data series can be derived [9] D. E. Echeverría, J.L. Rueda, D. G. Colomé, I. Erlich, “Improved
method for Real-Time Transient Stability Assessment of Power
from PMU measurements of voltages and currents, and this Systems”, IEEE Power & Energy Society General Meeting, San Diego,
information can be used to predict rotor angle at subsequent California, USA, July 2012.
time points. In view of this, this paper proposes an approach [10] I. Kamwa, A. K. Pradham, G. Joos, and S. R. Samantaray, “Fuzzy
that employs simple procedures for the prediction of rotor Partitioning of a Real Power System for Dynamic Vulnerability
Assessment”, IEEE Transactions on Power Systems, vol. 24, No. 3, pp.
angles and transient stability assessment from PMU data in 1356 – 1365, August 2009.
real-time. The prediction of generator rotor angle position at [11] D. Peña, Análisis de Datos Multivariantes, Editorial McGraw-Hill,
future points of time is performed by using the finite España, cap. 1 – 8.
[12] M. Pai, “Energy Function Analysis for Power System Stability”,
difference method. The simplified representation of a power Kluwer, Norwell, MA, 1989.
system by means of SMIB equivalent model allows [13] DIgSILENT Power Factory (2011). User’s Manual Version 14.1.
characterizing the transient stability in terms of the unstable Edición 1, DIgSILENT GmbH, Gomaringen, Germany.
[14] MATLAB, Math Works, Inc., Natick, MA, USA.
margin and the time to instability, which entails low www.mathworks.com
computational burden. Numerical results obtained by [15] M. Pavella, D. Ernst, and D. Ruiz-Vega, “Transient Stability of Power
applying the approach on the New England test system Systems: A Unified Approach to Assessment and Control”, Norwell,
Kluwer, 2000.
demonstrates the feasibility and effectiveness that could be
achieved in predicting the critical machines and unstable

You might also like