A Novel Real-Time Transient Stability Prediction Method Based On Post-Disturbance Voltage Trajectories

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/271496387

A novel real-time transient stability prediction method based on post-


disturbance voltage trajectories

Conference Paper · October 2011


DOI: 10.1109/APAP.2011.6180495

CITATIONS READS

6 48

6 authors, including:

Jinquan Zhao
Hohai University
59 PUBLICATIONS   529 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

distributed voltage stability assessment for integrated transmission and distribution grids View project

All content following this page was uploaded by Jinquan Zhao on 24 February 2016.

The user has requested enhancement of the downloaded file.


2011The International Conference on Advanced Power System Automation and Protection

A novel real-time transient stability prediction method based on


post-disturbance voltage trajectories
ZHAO JinQuan1*, LI Jun1, WU XiaoChen2, MEN Kun2, HONG Chao2 & LIU YongJun2
1
School of Energy & Electric Power Engineering, Hohai University, Nanjing 210098, China;
2
Electric Power Research Institute, China South Power Grid Ltd., Guangzhou 510623, China

Abstract˖With the wide application of phasor measurement units (PMU) and wide area measurement system (WAMS) in
power system, the real time detection and emergency control of power system transient instability based on the wide-area
measurements has brought wide attention. This paper presents a novel technique for predicting transient stability status of a
power system following a large disturbance. The synchronously measured samples of the fundamental frequency voltage mag-
nitudes are used to detect the stability. A stability criterion based on the volt-second integration of post-disturbance voltage tra-
jectory is proposed for detecting the transient instability. The drop rate of voltage magnitude after clearing the fault is embed-
ded into the proposed criterion since it indicates the severity of disturbance in some extent. There are two integration termina-
tion criteria for fast prediction. Compare with traditional transient stability prediction methods, in which the post-fault rotor
angle and angular velocity samples are used to produce the instability criteria, the proposed method improves the reliability
and rapidity of transient stability detection especially for the multiple swing instability cases. Studies with both IEEE 39 test
system and China Sothern Power Grid indicates that the proposed method is effective.

Keywords˖transient instability prediction, phasor measurement unit; post-fault voltage trajectory; volt-second inte-
gration criterion; voltage drop rate

1 Introduction power system is considered entering a transient instability


status. Reference [6] proposed the maximum rotor angular
The widely used special protection systems (SPS) provide velocity could be an indicator of transient instability. If it
control for potential single and multiple-related outages exceeds a predefined threshold value, then emergency con-
identified in the power system planning process. The setting trol should be enforced. The method and criteria are very
and strategy depend on off-line time-domain simulations simple and easy to be implemented. However, it has no
based on the models and parameters of power system. It can strict theoretical basis. For non-autonomous power system,
be called the feed-forward stability control. It is designed the threshold setting would vary with the operation envi-
for the given events. If an unexpected event happened, it ronments.
cannot do anything. With the wide application of GPS Reference [7] proposed a transient instability detection
technology and Phasor Measurement Unit (PMU) in power method based on the trajectory geometry characteristics in
system, the real time transient instability prediction based phase plane. Based on same idea, reference [8] developed
on the post-fault trajectories has brought wide attention the detection criterion into the angular velocity – angle
[1-5]. The scheme depends on real responses of power sys- phase plane while reference [9] further extended it into the
tem on a happening disturbance. It can be called the Feed- generator kenotic energy - angle phase plane.
back wide-area stability control. It measure power system According to the analysis and derivation of relationship
variables and can respond to arbitrary disturbances.If tran- between the curvature vector and transient stability, refer-
sient instability is detected, the emergency control actions ence [10] proposed that the curvature vector function could
such as cutting generator, cutting loads and system discon- be an indicator of the transient stability. The transient sta-
nection could be enforced within a short time. bility status can be detected rapidly. However, since it is
The real time measurements of transient instability de- belonging to the differential type criteria, it has a high re-
tection are limited to the post-fault rotor angle trajectories quirement for measurement quality. A small data error
for a long term. In references [1-5] the rotor angle differ- would induce incorrect stability prediction. References
ence between any two generators or one generator w.r.t. the [11-12] presented the extended equal area criterion (EEAC)
center of inertia COI angle reference are used as an indica- based real time transient stability prediction method. It uses
tor. If it exceeds a predefined threshold value, then the the power-angle integration area for the strictest equivalent
One Machine Infinite Bus (OMIB) as the stability criteria. It
APAP2011 www.apap2011.org
*Corresponding author (email: jqzhao2@tom.com)
___________________________________
978-1-4244-9621-1/11/$26.00 ©2011 IEEE

2011The International Conference on Advanced Power System Automation and Protection

can detect instability and calculate stability margins rapidly swing characteristics of the perturbed voltage trajectories,
and accurately. However, it depends on fast generator clus- the transient stability of power system can be detected by
tering technique and more data that are complete. It is diffi- observing the perturbed voltage trajectories.
cult to realize in a real large power grid.
In [13] Taylor et al. gave an algorithm used in their
¥ Severe perturbed generator
Wide-Area Stability and Voltage Control System (WACS)
in BPA. It based on 12 voltage magnitude measurements at
seven 500-kV stations in WECC. A weighted average vol-
tage could be computed from the 12 measurements, with Non-severe perturbed generators
highest weight for measurements close to where the voltage
swings are usually greatest. A nonlinear accumulator (inte-
t
grator) computes volt-seconds below a threshold setting
(525 kV for capacitor/reactor bank switching and 520 kV 
for generator tripping). Accumulation is blocked for voltage
Figure 1 Perturbed rotor angle trajectories.
recovery. If the volt-second accumulation reaches a setpoint,
then emergency control actions trigger; also, the weighted
voltage must be below 490 kV for generator tripping. Based Non-severe perturbed generators
on similar idea, [14] proposed a two-stage method which V
involves estimation of the similarity of post-fault voltage
trajectories of the generator buses to some pre-identified
templates and then prediction of the stability status using a
classifier which takes the similarity values calculated at the
different generator buses as inputs. The typical bus voltage Severe perturbed generator
variation patterns for both stable and unstable situations are
t
identified from a database of simulations. The same data-
base is used to train a support vector machine classifier 
which takes proximity of the actual voltage variations to the Figure 2 Perturbed voltage trajectories
identified templates as features.
This paper presents a novel technique for predicting tran-
sient stability. The perturbed voltage trajectories are used to
V
detect the stability status. A stability criterion based on the
volt-second accumulation of post-fault voltage trajectories
is proposed and the voltage drop rate after clearing the
faults is integrated since it can indicate the severity of fault
in some extent. Two integration termination criteria are de-
signed for stable cases and unstable cases respectively. The Critical instability
proposed method does not require the complete wide-area t
measurements and do not need to calculate the center of
inertia of the system. It improves the reliability and rapidity 
of transient stability detection especially for the multiple Figure 3. Different perturbed voltage trajectories with different disturbance
swing instability cases. Studies with both IEEE 39 test sys- severity
tem and China Sothern Power Grid indicates that the pro-
posed criterion and method are effective. According to figure1,2 and 3 obtained by a large amount
of off-line simulations, we could get the following observa-
2 Post-fault voltage trajectory characteristics tions about the perturbed voltage trajectories.
Observation 1: for most of short circuit faults, during the
The common characteristics of the post-fault voltage trajec- fault-sustaining phase, the bus voltage magnitudes have
tories can be used to indicate the disturbance severity of sudden drop and sudden return process. This phase of vol-
power system, which is the foundation of the post-fault vol- tage trajectory has no relationship with the final transient
tage trajectory-based transient instability prediction. instability. Therefore, in our method the volt-second inte-
Post-fault generator rotor-angle trajectories and bus voltage gration starts with the fault clearing time.
trajectories are shown in figure 1 and 2 respectively. It can Observation 2: for one certain contingency, the drop mag-
be seen that when a fault happened, with the swing of rotor nitude and drop velocity of voltage trajectories increase
angles of the severe perturbed generators, their terminal with the disturbance severity. The evolution rule of voltage
voltage magnitudes drop dramatically. According to the trajectories from stable to unstable is shown in figure 3.


2011The International Conference on Advanced Power System Automation and Protection

Observation 3: The swing magnitude of voltage trajectories lowest voltage value is used to integrate.
do not increase smoothly with disturbance severity. There is
3.3 The volt-second integration method
a salient jump for perturbed voltage drop between the criti-
cal stability case and the critical instability cases. There are two requirements for the voltage trajectory in-
tegration method. One is that the volt-second integration
value can indicate as accurately as possible the disturbance
3 Voltage trajectory integration criterion
severity. Another is that the voltage trajectories of stability
cases and instability cases can be distinguished by the inte-
Based on the perturbed voltage trajectories characteristics
gration value. The formula of the voltage trajectory integra-
shown above, we propose a transient instability detection
tion in [13] can be written as:
method, in which a integration value within the time domain temd
of a perturbed voltage trajectory is used as the indicator.  A ³ t0
[VU  V (t )]dt  (1)
3.1 The starting logic (disturbance detection) where t 0 is the integration beginning time, i.e. fault clear-
The proposed algorithm starts at the time of fault begin- ing time, t end is the integration end time, which is depend-
ning. It includes two phases: the disturbance detection pahse ing on the type of integration termination. We will describe
and the trajectory integration phase. At the moment of fault it later. V(t) denotes the voltage measurement of key bus at
happened, the voltage trajectories drop abruptly, power sys- time t . VU is the voltage upper limit of integration, here is
tem enter a phase of transient energy accumulation. At the
0.8 p.u.. The voltage upper limit is a threshold value for
moment of fault clearing, the voltage trajectories would
integrating the perturbed voltage trajectory. When the vol-
recovery abruptly, the power system enter a phase of tran-
tage trajectory swings below this value, the integration
sient energy dissipation. If the difference of two consecutive phase would start up. The voltage upper limit compressed
sampling voltage magnitudes exceeds a predefined thre- the scope of voltage integration for stable fault cases and
shold value (here is 0.2p.u.), then we can detect that a fault improved the sensitivity of the proposed criterion.
happened or a fault is cleared. If the voltage magnitude is By a large amount of off-line simulations, we analyzed
recovery, then a fault is cleared and the method enters the the swing characteristics of the perturbed voltage trajecto-
trajectory integration phase. It is necessary to note that if ries. According to the integration area of perturbed voltage
there are several continuous fault clearing actions, the vol- trajectory within first swing, since the drop rate of perturbed
tage trajectory integration value of the former action would voltage trajectory would increase with the severity of dis-
affect the detection of latter actions. Therefore, once the turbance for one contingency, at the same time the drop
fault clearing action is detected, the integration value would magnitudes are almost same for all instability trajectories.
be clear to zero and new cycle would be started. Therefore, for those unstable cases, the duration of voltage
Disturbance drop decrease with the increase of fault duration. The per-
detection turbed first swing integration area also decreases. Even
though it is still bigger than the integration area of stable
Integration
operation faults, the detection time is too long to take emergency con-
VU trol actions. Therefore, we use the voltage drop rate as the
coefficient in the proposed integration method. The integra-
tion formula becomes:
VD >V (t  W )  V (t )@ dt 
³ >V  V (t )@
tend
A1 (2)
t0 t t0 U
W
 where V(t W) is voltage magnitude at previous sampling
Figure 4 Schematic diagram of disturbance identification and integration time, W is the sampling time interval, A1 denotes im-
operation
proved voltage trajectory integration value. In this way, in
the process of voltage drop (from the peak to the valley
3.2 Determination of voltage trajectory integrated bottom), the integration value is positive, which denotes the
accumulation process of transient energy. In the phase of
The bus, which voltage magnitude is lowest at the mo-
voltage recovery (from valley bottom to next peak), the in-
ment of fault clearing (after voltage recovery) among all tegration value is negative, which denotes the dissipation
PMU measured buses, is considered as the key bus for this process of transient energy. The schematic diagram of our
disturbance. Its perturbed voltage trajectory is used as the integration method for a multiple swing instability case is
integration trajectory. In reality, the key bus may not be de- shown in figure 5. The integration value of the area above
termined and the low envelope curve of all measured bus the voltage upper limit is zero as shown in figure 5. 
voltage trajectories could be used as a combined voltage
trajectory. That is to say, at each measurement moment the


2011The International Conference on Advanced Power System Automation and Protection

integration operation would continue until the integration


V “+” area “-” area “0” area
value exceeds the given threshold value in a latter swing. It
VU can be seen in figure 5. The instability detection was given
in third swing.

VU
t0 tend t

Figure 5 Schematic diagram of integration method (multiple swing in-
stability)
t0 tend t

A large amount of off-line simulations show that if the 


Figure 6 Schematic diagram of integration termination (multiple
initial post-fault voltage magnitude is too low, the time and
magnitude of the voltage drop would decrease, which result swing stability)
in the integration value of first swing of perturbed voltage
trajectory cannot correctly indicate the disturbance severity. For the stable cases, when the voltage return to the vol-
Therefore, we propose that the integration value is modified tage integration upper limit and maintain a phase of time
by using the initial post-fault voltage magnitude. The inte- (here is 10 cycles), power system is stable and the integra-
gration formula becomes:
tion process is terminated. It is shown in figure 6.
VU  VD
A2 A1  (3) PMU mea- PMU mea-
V0  VD surements surements

where VD denotes the voltage lower limit, V0 denotes the


No Y No
initial post-fault voltage magnitude. The voltage lower limit Fault ? Fault clear?
is the lowest voltage magnitude of perturbed voltage trajec-
Phase 1: disturbance detection
tories for all instability faults cases. It is used to speed up Yes
the detection of the unstable faults. The purpose of design-
ing these voltage integration parameters is to distinguish the Determinate Integration
voltage trajectory
integration values between stable and unstable fault cases to
full extent.
Phase 2: Integration
PMU data
3.4 Integration termination criteria
Integration termination criteria include instability detec-
Integration
tion and stability detection. Since the integration value is the
function of the integration time, the transient instability cri-
terion is: Yes
Integration value
Unstable exceeds the
A2 ! Aset  (4) threshold value?

where Aset is a given threshold value. The determination of No


the threshold value requires to do a large amount of off-line Yes No
simulation for a large scale of contingencies with different Integration
Stable
end ?
types, different locations and obtain the perturbed voltage
trajectories, especially for those of critical stability and crit- 
ical instability. Since the sudden change characteristics of Figure 7. the flow chart of the proposed method
voltage trajectory from the stable cases to unstable cases,
we can get the critical integration value between stable tra-
In our method, the perturbed voltage trajectories are not
jectories and unstable trajectories by using the proposed
integration method above. The maximal integration value predicted by using the curve fitting methods. Since the vol-
among all stable contingencies is selected as the threshold tage measurements are used to detect, the system inertia
value of transient instability detection. center of angle reference is not estimated. Instead, the bus at
For multiple swing instability cases, the integration val- which the voltage drop rapidly need to be identified quickly.
ue would not exceed the given limit in the process from the Our method has better data fault tolerance. Beneficially,
fault clearing time to the valley of first swing. However, the faster operation results for more severe disturbances.


2011The International Conference on Advanced Power System Automation and Protection

4 Implementation and the flow chart of our me-


thod

The main steps of the proposed method are as follows:


Step 1: Obtain the voltage measurements (trajectories) by
PMU/WAMS and detect the disturbance (fault occurs and
fault clears);
Step 2: Determine the voltage trajectory used for integration
and integrate the perturbed voltage trajectory using formu-
las (2) and (3);
Step 3: When the integration value exceeds the given thre-
shold value at a certain time, then detect power system un- (A) fault duration 0.13s
stable and enforce the emergency control actions such as
cutting off the most perturbed generators. If the voltage tra-
jectory returns above the voltage upper limit and maintain a
phase of time, then detect power system stable and termi-
nate the volt-second integration.
The flow chart of our perturbed voltage trajectory based
transient instability detection method can be seen in figure
7.

5 Results and discussion 


(B) fault duration 0.14s
Both the IEEE39 bus system and China South power grid
are used to testify our proposed method. The perturbed vol-
tage trajectories come from BPA and PSASP simulations,
the sampling time interval is 0.01s.
IEEE39 bus system
At first, all 10 generator terminal voltages in IEEE 39
bus system are selected as the PMU measurements. In order
to testify the validity of our method, we obtain the perturbed 
voltage trajectories by PSASP simulations and compare the (C) fault duration 0.15s
proposed method with reference [13]’s method and the rotor
angle difference threshold value method.
We set a three phase short circuit fault happened at
branch 26-29 and the fault durations are 0.13s, 0.14s, 0.15s
and 0.2s respectively, which correspond to the cases of sta-
bility, multiple swing instability and single swing instability.
The perturbed voltage trajectories are given in (A), (B), (C)
and (D) of Figure 8. The modification results of the integra-
tion value by using the proposed method are given in table 1.
It can be seen in figure 1, the integration value of first swing 
of perturbed voltage increases with the disturbance severity (D) fault duration 0.20s
(here is actually the fault duration) by using our improved Figure 8.Perturbed voltage trajectories for 3 phase short circuit at branch
voltage trajectory integration method, which improve the 26-29
availability and practicality of our voltage trajectory inte-
gration method. The integration value for case of fault dura-
Table 1 Comparison of the integration values by using two methods
tion 0.20s is smaller than that of fault duration 0.15s by
[13]’s method. Fault line Fault Duration [13]’s method Proposed me-
thod
26-29 0.13s 0.0085 0.0020
26-29 0.14s 0.0359 0.0111
26-29 0.15s 0.1974 0.1894
26-29 0.20s 0.0940 0.2070


2011The International Conference on Advanced Power System Automation and Protection

The threshold value is set as 0.01. The stability detection of power grid. However, still some unexpected events hap-
results with different methods are shown in table 2. It can be pened in each year.
seen in table 2, our improved voltage trajectory integration We selected 35 buses as the voltage phaosr measurements.
method can detect accurately the transient instability. Com- They include generator terminal buses and some 500kV
pare with the threshold value method of rotor angle differ- buses. The tested contingencies include single contingencies
ence, our method has a large improvement in detection time, and complex contingencies. The contingency descriptions
which gives more time margin for transient stability emer- are given in table 4 and their detection results are shown in
gency control. For case of fault duration 0.14s which ac- table 5. 
tually is an unstable case, the EEAC method give wrong
result and judge it a stable case. In addition, compare with Table 4 fault case description of china south power grid
the real time measurement based EEAC method, our me-
thod does not require predicting the perturbed trajectories Case Fault description
and has easier computation process. The required stability 1 3 phase short circuit at line LP-BS (0.34s duration)
detection time is less than the real time measurement based 2 3 phase short circuit at line LP-BS(0.36s duration)
EEAC method. Furthermore, our method can give accurate 3 3 phase short circuit at line LP-BS(0.40s duration)
detection result on multiple swing instability cases. 4 bipolar block in Yun-Guang HVDC (0.1s duration)
5 3 phase short circuit at line Anqing I, Anshun side break-
er failure and switch off line Angui I
Table 2 Comparison of the stability detection time by using different me-
thods
Table 5 the detection results for China south power grid
Fault line Fault Proposed Measurement 180° angle
Case Proposed [13]’s method Rotor angle
duration method based EEAC difference
method difference 250°
26-29 0.13s 0.3s Stable 0.53s stable -
1 stable stable -
26-29 0.14s 0.58s 0.64s stable 10.72s
2 1.48s 1.54s 1.39s
26-29 0.15s 0.44s 0.74s 1.17s
3 1.15s 1.34s 1.26s
26-29 0.20s 0.32s 0.38s 0.71s
4 1.42s 1.62s 1.95s
For case of branch 13-14 three-phase short circuit and 5 0.35s failed 0.80s
fault duration 0.21s, we try to detect the transient stability It can be seen in table 5, reference [13]’s method failed
status by using same threshold value (0.01). In addition, in for case 5 while the proposed method can give correct sta-
order to testify the detection effectiveness of hybrid mea- bility detection results. The detection time are less than both
surements (not only the generator terminal buses, but high [10]’s method and rotor angle difference threshold value
voltage side buses), we suppose that generator terminal bus method. 
31 has no voltage phasor measurement, in its place it is the
voltage of bus 6. The numerical results are shown in table 3. 4 Conclusion

Table 3 Comparison of the detection time by using pure generator terminal A novel method for fast detecting transient stability sta-
voltage and hybrid measurements tus after a severe disturbance is presented in this paper. The
Measurement type Proposed me- 180° angle synchronously measured samples of the fundamental fre-
thod difference quency voltage magnitudes are used to replace the rotor
Generator terminal voltage 0.47s 1.29s angle measurements. The proposed stability criterion is
based on the volt-second accumulation of post-fault voltage
Hybrid measurement 0.53s 1.29s
trajectories. The voltage drop rate after clearing the faults is
It can be seen in table 3, the stability detection results are integrated since it can indicate the severity of fault in some
same before and after the hybrid measurements are used. extent. Two integration termination criteria are designed for
Therefore, in real application when not all of generators stable cases and unstable cases respectively. The proposed
installed PMU (i.e. not all of their perturbed voltage trajec- method does not require the complete wide-area measure-
tories are obtained), they can be replaced by their high vol- ments and do not need to calculate the center of inertia of
tage side buses. the system. It improves the reliability and rapidity of tran-
sient stability detection especially for the multiple swing
5.2 China South power grid instability cases. The detection time is less than other me-
China South power grid is one of biggest DC-AC hybrid thods. Numerical simulation results with both IEEE 39 bus
power grid in the whole world. Its dynamic characteristics system and China Sothern Power Grid indicate that the
are very complicated. More than 10 special protections sys- proposed method is effective.
tems are installed for some given events in planning phase


2011The International Conference on Advanced Power System Automation and Protection

This work was supported by the National Natural Science Foundation of 7 Wang L, Girgis A A. A new method for power system transient stabil-
China (Grant No. 51077042) and the Key Foundation of China South Power ity detection. IEEE Trans. On power delivery, 1997, 12(3): 1082~
Grid. 1089.
8 Xie H, Zhang B H, LI G. Real-time prediction of transient instability
1 Liu C W, Thorp J. Application of synchronized phasor measurements based on wide-area information of generator state. Electric Power
Automation Equipment, 2009, 29(7): 28-32. (in Chinese)
to real-time transient stability prediction. IEE Proc. Gener. Transm.
Distrib. 1995, 142(4): 355~360. 9 Xie H, Zhang B H, Yu G L. Transient instability detection based on
2 Rovnyak S, Liu C W, Lu J, Ma W, Thorp J. Predicting future beha- trajectory geometrical characteristic. Proceedings of the CSEE, 2008,
28(4): 16-22. (in Chinese)
vior of transient events rapidly enough to evaluate remedial control
options in real-time. IEEE trans. On power systems. 1995, 10(3): 10 Song F F, Bi T S, Yang Q X. Study on WAMS based on multi swing
1195 ~1203. stability assessment for power systems. Proceedings of the CSEE,
2006, 26(16): 38-45.(in Chinese)
3 Liu C W, Thorp J S. New methods for computing power system dy-
namic response for real-time transient stability prediction. IEEE 11 Teng L, Liu W S, Yun Z H. Study of Realtime Power System Tran-
Trans. On Circuits and Systems: Fundamental Theory and application, sient Stability Emergency Control. Proceedings of the CSEE, 2003,
23(1): 64-69. (in Chinese)
2000, 47(3): 324~337.
4 Daoud A, Karady G, Amin A. A new fast-learning algorithm for pre- 12 Zhang P F, Xue Y S. Quantitative Transient Stability Assessment
dicting power system stability. Proceedings of 2001 IEEE PES winter Method Using Phasor Measurement. Automation of Electric Power
Systems, 2004, 28(20): 17-20. (in Chinese)
meeting, 2001, Jan. 28-Feb. 1, Singapore: 594-598.
5 Takahashi M, Matsuzawa K, Sato M, et al. Fast generation shedding 13 Taylor C W, E Dennis C. Erickson, Kenneth E. Martin.
equipment based on the observation of swings of generators. IEEE WACS-Wide-Area Stability and Voltage Control System: R&D and
Online Demonstration. Proceedings of The IEEE, 2005, vol. 93, no.5,
Transactions on Power Systems, 1988, 3(2):439-446.
6 Stanton S E, Slivinsky C, Martin K, et al. “Application of phasor pp.892-906.
measurements and partial energy analysis in stabilizing large distur- 14 Athula D Rajapakse, Francisco Gomez, Kasun Nanayakkara. Rotor
Angle Instability Prediction Using Post Disturbance Voltage Trajec-
bances” IEEE trans. On power systems, 1995, 10(1): 297-306.
tories. IEEE Transaction On Power Systems, 2010, vol. 25, no. 2, pp.
947-956.



View publication stats

You might also like