You are on page 1of 8

2846

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 29, NO. 6, NOVEMBER 2014

Preventive Control Stability


Via Neural Network Sensitivity
Mauricio C. Passaro, Member, IEEE, Alexandre P. Alves da Silva, Senior Member, IEEE, and
Antonio C. S. Lima, Member, IEEE

AbstractThis paper discusses the power systems stability


margin improvement by means of preventive control based on
generation re-dispatch using a neural sensitivity model. This model
uses multilayer perceptron networks with memory structure in the
input layer. The training of this model is made with temporal data
samples from time domain simulations, incorporating information
about the dynamic behavior of the system, unlike the methods
proposed in the literature in which the pre-fault system data are
used instead. The sensitivity is used as a guideline in selecting
the most effective set of generators in the reallocation of the
amount of active power capable of increasing system security. The
effectiveness of the proposed methodology has been demonstrated
through the application to a large system.
Index TermsNeural network application, power system dynamic stability, power system security.

I. INTRODUCTION

LECTRIC power systems are strategic for the economic


activity and its constant evaluation is very important for
ensuring continuity of energy supply to society. In recent years,
these systems have gone through several changes. The process
of deregulation of the electricity industry in many countries
has brought new challenges for the interconnected systems operation. As they were operating close to their limits [1], this
new scenario led to the occurrence of several blackouts worldwide. For instance, in 1996 a blackout affected the West Coast
System (WSCC) [2], and in 1999 [3] and 2009 [4] a system wide
blackout affected most of the Brazilian National Power Grid.
The recurrence of these major failures emphasizes the importance of an efficient restoration procedure [5] and the need for
the development of fast, secure and reliable security assessment
tools. Transient stability is without a question a critical problem,
although solution via analytical techniques alone does not allow
preventive or corrective actions in a timely manner.

Manuscript received August 22, 2013; revised December 02, 2013 and March
21, 2014; accepted March 29, 2014. Date of publication April 10, 2014; date of
current version October 16, 2014. This work was supported in part by INERGE,
FAPEMIG, CNPq, and CAPES. Paper no. TPWRS-01085-2013.
M. C. Passaro is with the Department of Electrical Engineering, Federal
University of Rio de Janeiro, Rio de Janeiro, RJ, Brazil, and also with the
Brazilian Independent System Operator (ONS), Rio de Janeiro, RJ, Brazil
(e-mail: mcp@ons.org.br).
A. P. A. da Silva is with GE Global ResearchBrazilian Technology Center,
Rio de Janeiro, RJ, Brazil (e-mail: dasilva@ge.com).
A. C. S. Lima is with the Department of Electrical Engineering, Federal University of Rio de Janeiro, Rio de Janeiro, RJ, Brazil (e-mail: acsl@dee.ufrj.br).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TPWRS.2014.2314855

Transient stability is, according to many authors [6], the most


studied problem in power systems dynamic performance. There
is a vast amount of publications [7] dedicated to improve transient stability and they have in common the use of active power
redispatch to achieve this goal [8]. The main issue is to determine the amount of generation to be relocated and the selection of generators. The integration of computational intelligence
(CI) techniques with analytical methods can provide significant
improvement in the security assessment process, given that stability studies databases available in electric utilities planning departments contain useful information.
Nowadays, the strategic planning of the Brazilian Independent System Operator (ONS) has focused on the implementation and operation of a computational tool to analyze static and
dynamic security in real time [9]. Based on the results of state estimation or based on data from short-term planning studies, the
process is to determine the conditions under which the system
meets the criteria established by the operating procedures [10].
The results obtained by simulation and presented in the form
of security regions or nomograms have been helping planning
and operation teams to streamline the analysis for different operating situations and to support decision making from real-time
operation teams [11]. In this context, it is interesting to integrate this analysis tool with CI techniques, specifically neural
networks, to provide a set of rules and operational guidelines for
preventive control, in order to improve the dynamic behavior of
the system after redispatch.
This paper describes a methodology based on an application of neural networks (NN) for the determination of a sensitivity model between stability margins and generation redispatch. The NN are trained with data from electromechanical
transient simulations, including information on the dynamic behavior of the system, unlike almost all the previous methods presented in the literature [12]. The few exceptions have only dealt
with small-scale problems [13]. Another highlight is the development of a control strategy based on the proposed sensitivity
model. Although inaccuracies in models or field measurements
might affect the results, at this point is not certain how to assess these issues and their impact on the system performance.
This topic is considered open and left for future research. Nevertheless, the models used in the transient stability studies for
the Brazilian system have been able to reproduce large blackouts as well as some minor events [3], [5].
The results encourage the use of the proposed model in the
real time operation environment, with the objective of generating preventive control rules based on the dynamic behavior
of the system. These rules may be part of the operation guidelines of the Brazilian Independent System Operator. In order to

0885-8950 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

PASSARO et al.: PREVENTIVE CONTROL STABILITY VIA NEURAL NETWORK SENSITIVITY

2847

apply the proposed approach to the real-time operation environment, a base load-flow can be generated from forecasted data.
The whole analysis can be performed considering the estimated
load and the generation schedule for some minutes ahead [14].
II. PREVENTIVE CONTROL
Traditionally, operational security control is divided into two
main categories: preventive control and emergency control.
The purpose of preventive control is to prepare the system
when it is still in normal operation, so as to make it capable of
facing possible future events satisfactorily. On the other hand,
in emergency control, disturbances have already occurred, and
the goal is to control the dynamics of the system to mitigate the
consequences.
Regarding the type of control action, the preventive mode is
implemented by generation redispatch, switching network components such as compensation elements or transmission lines, or
by load shedding. In emergency control, the actions are limited
to load or generation shedding, switching shunt capacitors/reactors, or network partitioning, coordinated by means of special
protection schemes.
Preventive control applies sensitivity analysis as the basic
mechanism for decision making. Energy margin analytical sensitivity [8], [15], [16] involves a high computational cost. One
alternative is based on neural networks [12]. With neural networks two ideas have been studied: 1) The use of neural networks in transient stability assessment with the purpose of classification [17] and preventive control via another procedure,
such as generation redispatch and/or load shedding based on
optimization or using decision trees [15]; 2) The use a neural
sensitivity model [12], [19][21].
The sensitivity evaluation by a neural model provides the partial derivative of a mapping with no explicit analytical formulation, and it is dependent on a large data set of systemic information. Note that neural networks present desirable characteristics
such as fast response, simplicity in its output format (stable/unstable or stability margin), and flexibility in managing uncertainty. The high computational burden during the training phase
is performed off-line in a planning environment.
The proposed neural sensitivity model uses the stability
margin calculated by time domain simulations as an index of
systemic security [11].
A. Stability Margin
The power system transient stability analysis can be performed using the concept of energy and security margin [9].
Electric power system transient instability is characterized by
separation of the system into two parts, i.e., a group of units
called critical machines, which distances itself from the rest
of the generation system (non-critical machines). This feature
allows the system stability study using the equal area criterion, where the two groups are represented by two equivalent
machines.
Therefore, the power system transient stability can be evalusing the foluated using the concept of stability margin
, the system is considered stable from the
lowing criteria:
standpoint of transient stability;
, the system is considered
unstable from the standpoint of transient stability.

Fig. 1. Power angle curve.

If the system is stable, the stability margin


from the following equation:

is calculated

(1)
If the system is unstable (acceleration area A1 larger than the
deceleration area A2), the negative margin is numerically equal
, according to Fig. 1. The
to the kinetic energy at the point
value of this margin is calculated by [9].
Considering a list of pre-defined contingencies, the stability
margin of the system must satisfy the following relationship
[20]:
(2)
where is the stability margin and
is the minimum stability
margin. If the electrical system is in an unsafe operation condition with respect to contingency , control actions should modify
the stability margin, satisfying the following relationship:
(3)
where is the stability margin related to the i-th contingency.
The change required on the stability margin is estimated through
the sensitivity coefficient:
(4)
where
is the sensitivity of the stability margin with recorrespect to the control variable P (active power), and
sponds to the control variable change.
III. NEURAL NETWORK SENSITIVITY MODEL
The proposed sensitivity model uses a conventional multilayer perceptron network with a memory structure in its input
layer. The training, validation and test sets consist of temporal
sampled data obtained from time domain simulations.
A. Neural Network
Among the applications of neural networks in power systems, the great majority use the multilayer perceptron architecture [22]. According to this publication, about 400 technical articles on the theme were investigated, and 81.19% of the total
has used multilayer perceptrons.

2848

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 29, NO. 6, NOVEMBER 2014

Fig. 2. Data acquisition intervals.

B. Temporal Data Sample


Predominantly, both transient stability assessment and preventive control based on neural sensitivity model use pre-fault
system data to generate training sets for the neural network
[19][21], [23], [24].
In [18] and [22], the use of dynamic data was discouraged. In
[24], it is reported that there is insufficient evidence to determine
whether the static variables are more appropriate than dynamic
variables (temporal data samples) as attributes for neural networks applied to power systems transient stability analysis. The
neural models proposed by [25][27] use temporal data samples
from time domain simulations as input variables for feeding the
network. However in these papers the neural model aims to classify the system for stability/instability only.
The proposed neural sensitivity model uses temporal data
samples for build the training, validation and testing sets, as
shown in Fig. 2. The model output is the stability margin for
the operation point represented by the selected input variables.
According to Fig. 2, the sampling is done considering three
distinct intervals. The first interval (P1) contains samples obtained during the application of a 110 ms fault. The interval (P2)
contains information between 50 ms after the beginning of the
fault and 100 ms its extinction, i.e., with a duration of 150 ms.
The interval (P3) contains post-fault values and ranges from 250
to 750 ms. The purpose of dividing the data sampling is the verification of the period that results in a better sensitivity neural
model, capable of generating active power reallocation rules for
improving the stability margin.
C. Memory Structure
A simple way to insert short-term memory in the neural structure is the insertion of time delay on the input variables. Fig. 3
shows a memory structure representation applied to the neural
network input layer. Fig. 4 shows the importance of the memory
structure representation in a multi-layer perceptron network.
The 9 buses test system proposed in [28] is used, in which four
contingencies were simulated. Each contingency has a stability
MW,
MW, 14
margin for a given operating point of
MW, and 19 MW, respectively.
The stability margin estimated by the neural model without a
memory structure in its input layer (star markers) shows a large
dispersion around the true margin (calculated by using conventional time domain simulation). Including short-term memory

Fig. 3. Time delay in the input layer.

Fig. 4. Stability margin estimation.

in the neural model provides an increase in accuracy for estimating the stability margin.
D. Input Variable Selection
The selection of the input data of neural networks applied to
transient stability assessment problems, traditionally, is based
on the experience of experts. Such entries typically describe
adequately the power system state and the stability evaluation
from these attributes is conservative. The dimensionality of
the problem should be as small as possible in order to reduce
the computational burden and improve the performance of the

PASSARO et al.: PREVENTIVE CONTROL STABILITY VIA NEURAL NETWORK SENSITIVITY

2849

neural model, however without deteriorating its generalization


capability.
While linear analysis methods (such as correlation) are useful
in particular cases, it is essential to consider non-linear relationship between different variables. The calculation of mutual information allows the assessment of non-linear dependence between them [28].
The mutual information for continuous systems is defined:
(5)
where p is the probability for the possible valuations of the
variables x and y. Tests were performed using the New England 39-bus test systems. These tests indicated that the variables most relevant to the stability margin are electrical power
. The latter is treated as a non-control(Pe) and rotor angle
lable variable while the former is considered as a controllable
one. Several compositions of input variables (controllable and
non-controllable, e.g., reactive power, accelerating power and
kinetic energy) were used in the neural network training. It was
observed that the use of electric power (Pe) and rotor angle
leads to better results. Also indicated that better sampling period of temporal data is the post-fault (P3). Probabilities density
functions in (5) have been transformed to the discrete domain
using histograms.
The use of the respective rotor angles has the objective of
evaluating the neural model performance impact in the training
process and sensitivity evaluation phase. In this case the inputs
of the neural network for the uncontrolled variables will not be
stimulated and control rules will be obtained only through the
changes imposed on the entries for the controllable variables.

Fig. 5. Block diagram of sensitivity modelRules module.

E. Sensitivity Analysis
The interdependence between the power system variables can
be quantitatively determined by sensitivity analysis. Sensitivity
is defined as the ratio of change on the dependent variable because of an independent variable variation. This analysis is very
important in power systems operation planning studies. It helps
in observing the cause-effect relationship, providing the basis
for system control actions evaluation.
In [5] and [10], analytical expressions to obtain the energy
margin sensitivity with respect to system parameters were applied. In [11], the sensitivity was calculated using the Taylor series expansion of the stability margin function. In [19], a neural
sensitivity model that employs a back-propagation algorithm
with fuzzy controller and activation function optimization was
proposed. This model uses pre-fault system data to generate the
training, validation and testing sets.
In the present work, model output (stability margin) is evaluated after variations of a given input, keeping the remaining
ones unchanged. The changes consist of positive and negative
steps of 10, 20 and 30% on the current input value.
F. Modules
The neural model construction process is divided into two
stages. The first stage involves the entire systemic data preparation, generation of analysis scenarios, variable selection, network training and the creation of preventive control rules. This

Fig. 6. Conceptual structure of sensitivity modelOperation module.

stage is executed in planning environment. The corresponding


flowchart is depicted in Fig. 5.
At the end of the process described in Fig. 5, the sensitivity
neural model is ready to be used in a real time environment.
The use of the neural model in a real time environment includes
the following steps: 1) systemic data acquisition via state estimation; 2) dynamic simulations for a contingency list defined
in the planning environment; 3) temporal data sampling for the
selected variables; 4) database building; 5) feed operation point
to the neural model; 6) if the margin obtained by the model is
negative, proceed to redispatch in accordance to the rules established in the planning stage (see Fig. 6).
The set of rules for preventive control is obtained based on the
sensitivity analysis results. For each geo-electrical area, a rank
of sensitivities is obtained in order to prioritize the generation
units to be re-dispatched. The most effective generation unit for
stability margin improvement when the corresponding output
is increased gets first place. Second place goes to the unit that
improves the stability margin the most while it has its generation
decreased. These pair of units are the ones to be re-dispatched,
keeping the generation-load balance. In the real-time operation
of the Brazilian Independent System Operator this information
is quickly obtained using a distributed processing hardware.

2850

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 29, NO. 6, NOVEMBER 2014

Fig. 8. NomogramOperation points.

TABLE I
OPERATION POINTS AND STABILITY MARGINS

Fig. 7. Brazilian National Interconnected SystemEquivalent.

IV. APPLICATION
In order to demonstrate the applicability of the proposed
methodology, the Brazilian National Interconnected System
(SIN), a large scale power system has been used. The system
consists of 4077 buses, 260 power plants, 3695 transmission
lines, 2146 transformers, 13 SVC, and 3 HVDC systems (Itaipu,
Garabi, and Alumar). Fig. 7 presents a simplified diagram of the
interconnections North-East, North-South, and East-Northeast.
In all operation points, a single contingency, three-phase
fault, involving one of the tie lines between the North/Northeast
is considered. The fault clearing time is equal to six cycles
(100 ms). The bus on which the fault is applied is Ribeiro
Gonalves, causing the opening of the Ribeiro Gonalves/Colinas and Ribeiro Gonalves/So Joo do Piau transmission
lines. The operating point of the base case corresponds to a
medium load condition. The power flow in the tie lines is 3100
MW (Northeast area is importing).

A. Training Data Generation


The creation of the database from which the training, validation and testing sets are extracted was performed by calculating
dynamic security regions, a feature of Organon [9]. The use the
nomograms derived from these calculations helps in the preparation of the operating points used for generating training, validation and testing data for the model.
The major advantage of this procedure is the possibility of visual interpretation of critical regions, which allows one to focus
the model synthesis on the regions around the stability border
[29]. Operating points used in the preparation of data sets for
training, validation and testing of the neural model can be seen
in Fig. 8. Table I shows the stability margins associated to each
operating point.

B. Variables Selection
The variables selection process of the Brazilian Grid aims to
indicate the list of more relevant generation units (with minimal redundancy) regarding stability margin. Initially, 76 power
plants are pre-selected, and the corresponding electrical powers
are analyzed. The initial set of 76
(Pe) and rotor angles
power plants is defined according to the criteria below: 1) operative reserve exceed 90 MW (because the margin of the operation
MW); 2) unit must be interconpoint used in the tests is
nected to the system through the main grid (230 kV and above).
The previous selection of generators with operative reserve
ensures that only units without technical limits can compose the
list of generators for which sensitivity will be assessed by the
neural model.
The variable selection process consists of two phases. The
first phase verifies which variables have higher relevance with
respect to stability margin. The second phase reduces the redundancy between the previously selected variables. Fig. 9 shows
the input selection process flowchart. Table II shows the variables ranking according to the mutual information value regarding stability margin.
It can be observed a variation of 24.21% between values
for buses 4520 (P.Pedra) and 6420 (Tucuru I), which corresponds to the greatest change in the sorted MI list. Therefore,
the more relevant variables are selected above that a cutoff. The
power plants that precede Tucuru I (6420) are discarded, with
remaining power plants totalizing 29.
After that, the redundancy minimization scheme described in
Fig. 9 is applied. The final list of power plants can be seen in
Table III. The list of selected power plants saves 11.84% of the
original list. In order to allow an understanding of the spatial distribution of the selected units, they are presented in Table III according to the corresponding geographical regions (north, northeast, and southeast).

PASSARO et al.: PREVENTIVE CONTROL STABILITY VIA NEURAL NETWORK SENSITIVITY

2851

TABLE IV
PAIRS OF POWER PLANTS FOR REDISPATCH

The neural network training, validation and testing sets were


formed from operating points show in Fig. 8. These operation
points led to temporal data samples totalizing 2060 patterns.
V. RESULTS
The model is validated by comparing its performance in assessing sensitivity with time domain simulations, in which the
possible alternatives for relocation are verified.
A. Time Domain Simulation

Fig. 9. Relevance maximization and redundancy minimization.

TABLE II
MUTUAL INFORMATION VALUES

TABLE III
SELECTED POWER PLANTS

C. Neural Network Training


The Matlab Neural Network Toolbox has been applied, and
the learning parameters have been adjusted according to the experience. The sampling rate is 5 ms and temporal data samples
are extracted after the fault (P3).

According to the nomogram shown in Fig. 8, the stability


MW. It is
margin for the operating point used in the test is
observed that power plants of the northeastern region should increase their generation at least 90 MW for the system to operate
in a secure condition. This power generation increase should be
compensated by decreasing generation in the north or southeast
power plants. A total of 100 MW of generation for relocation is
established. The list of selected power plants (Table III) establishes combinations of power plants that can be used as controls.
The pairs of power plants list can be observed by the Table IV.
The cells highlighted in red indicate that the preventive control via redispatch using these power plants is not effective, i.e.,
does not lead the system to a secure operating point. These pairs
belong to the same generation group or do not involve power
plants of the northeastern area.
The effective redispatch is compared by monitoring the
systemic kinetic energy as shown in Fig. 10. The pairs of
power plants leading to lower kinetic energy are 5061-20,
5061-754, 5061-3964, and 5061-36. The plants Marimbondo
(20), Cachoeira Dourada (754), Cana Brava (3964), and Serra
da Mesa (36) belong to southeast region. Among the plants
in the Northeast, Xing represents the best redispatch option
followed by Sobradinho, Paulo Afonso I, and Apolonio Sales
. The sorted list according
to the maximum kinetic energy system obtained by Fig. 10 is
shown in Table V.
B. Neural Model
The neural network topology presents 54 inputs (3 9 control variables with corresponding angles), 27 neurons in the
hidden layer and one output. This topology has been obtained
empirically through several tests, in which the numbers of neurons and hidden layers has been modified.
In fact, cross validation has made the estimated test error quite
immune to the number of neurons in the hidden layer. Fig. 11
shows the neural sensitivity model output.
It can be seen that a generation increase in Xing (bus 5061)
and a generation decrease in Marimbondo (20), Cachoeira
Dourada (754), Cana Brava (3964), and Serra da Mesa (36)
improve the stability margin (tendency to make it positive).

2852

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 29, NO. 6, NOVEMBER 2014

Fig. 10. Ranking according to kinetic energy (Vke).

TABLE V
PRIORITY PAIRS FOR REDISPATCH

Fig. 12. Sensitivity assessmentPower Plants from Northeast.

Fig. 13. Sensitivity assessmentPower Plants from North (Tucur and Xing).

Fig. 11. Sensitivity assessmentXing x SE Power Plants.

The generation increase in the Northeast implies reducing the


generation of North or Southeast areas. The sensibilities represent the expected behavior and it can be establish an order of
.
priority for redispatch:
The model output indicated to use only power plants in the
Northeast, as shown in Fig. 12. The list of priority obtained is:
. In Fig. 13 is depicted the
low sensitivity presented by Tucuru (6425). As indicated by
the time-domain simulation, this power plant had little effect on
the stability margin improvement.
VI. CONCLUSION
This paper proposes a methodology for determining a sensitivity model between stability margin and generation redispatch.
This methodology is based on neural networks trained with temporal data samples originated from transient simulations. The
adopted neural network is a multilayer perceptron with memory
structure in the input layer.
The use of temporal data samples and the memory structure
in the neural network input layer provides a model with accurate

stability margin estimation capability, and capacity for ranking


candidates for redispatch by sensitivity analysis.
The proposed input variable selection procedure has shown
ability to tackle the problem of high dimensionality, inherent to
large power systems. The application of this procedure to the
Brazilian National System has proven its effectiveness.
The developed preventive control rules are appropriate and
have been validated by time domain simulations, where the
redispatch considering pairs of power plants has been evaluated. Highly probable contingencies are already included
in market-based decisions. For less probable contingencies,
preventive control would not align with market decisions. The
proposed approach can provide useful information with respect
to the trade-off between cost and system vulnerability.
REFERENCES
[1] A. P. A. da Silva, Overcoming limitations of NNs for on-line DSA,
in Proc. IEEE PES General Meeting, San Francisco, CA, USA, Jun.
2005.
[2] C. Y. Chung, L. Wang, F. Howell, and P. Kundur, Generation
rescheduling methods to improve power transfer capability constrained by small-signal stability, IEEE Trans. Power Syst., vol. 19,
no. 1, pp. 524530, Feb. 2004.
[3] P. Gomes, New strategies to improve bulk power system security:
Lessons learned from large blackouts, in Proc. 2004 IEEE Power Engineering Society General Meeting, Denver, CO, USA, Jun. 2004, vol.
2, pp. 17031708.
[4] ONSOperation Daily Report, Nov. 10, 2009 [Online]. Available:
http://www.ons.org.br

PASSARO et al.: PREVENTIVE CONTROL STABILITY VIA NEURAL NETWORK SENSITIVITY

[5] P. Gomes, A. C. S. Lima, and A. Guarini, Guidelines for power system


restoration in the Brazilian system, IEEE Trans. Power Syst., vol. 19,
no. 2, pp. 11591164, May 2004.
[6] M. R. Aghamohammadi, A. Maghami, and F. Dehghani, Dynamic
security constrained rescheduling using stability sensitivities by neural
network as a preventive tool, in Proc. IEEE PSCE Power System Conf.
Expo., Seattle, WA, USA, Mar. 2009, pp. 17.
[7] D. Z. Fang, Y. Xiaodong, and S. Jingqiang, An optimal generation
rescheduling approach for transient stability enhancement, IEEE
Trans. Power Syst., vol. 22, no. 1, pp. 386394, Feb. 2007.
[8] K. N. Shubhanga and A. M. Kulkarni, Stability-constrained generation rescheduling using energy margin sensitivities, IEEE Trans.
Power Syst., vol. 19, no. 3, pp. 14021413, Aug. 2004.
[9] J. L. Jardim, Online dynamic security assessment: Implementation
problems and potential use of artificial intelligence, in Proc. IEEE
PES Summer Meeting, 2000, vol. 1, pp. 340345.
[10] Brazilian ISO, Procedures of Operation Manual, In: Grid Procedures,
Module 10, Submodule 10.21, Rev 1.1, 2010 [Online]. Available:
http://www.ons.org.br
[11] J. L. Jardim, C. A. S. Neto, and M. G. Santos, Brazilian system operator online security assessment, in Proc. IEEE PES Power Systems
Conf. Expo., Atlanta, GA, USA, Oct. 29, 2006, pp. 712.
[12] J. N. Fidalgo, J. A. P. Lopes, and V. Miranda, Neural networks applied to preventive control measures for the dynamic security of isolated power systems with renewables, IEEE Trans. Power Syst., vol.
11, no. 4, pp. 18111816, Nov. 1996.
[13] M. Djukanovic, D. J. Sobajic, and Y. H. Pao, Neural-net based calculation of voltage dips at maximum angular swing in direct transient
stability analysis, Int. J. Elect. Power Energy Syst., vol. 14, no. 5, pp.
341350, 1992, Elsevier Science.
[14] J. L. Jardim et al., A unified online security assessment system, in
Proc. CIGRE Biennial Session, Paris, France, Aug. 2000.
[15] V. Vittal, E. Z. Zhou, C. Hwang, and A. A. Fouad, Derivation of stability limits using analytical sensitivity of the transient energy margins, IEEE Trans. Power Syst., vol. 4, no. 4, pp. 13631372, Nov.
1989.
[16] C. R. Minussi and W. Freitas, Sensitivity analysis for transient stability, IEE Proc. Gen., Transm., Distrib., vol. 145, no. 6, pp. 669674,
1998.
[17] B. D. A. Selvi and N. Kamaraj, Transient stability assessment using
fuzzy SVM and modified preventive control, Int. J. Elect., Comput.,
Syst. Engineer, vol. 2, no. 1, p. 64, Mar. 2008.
[18] H. Vasconcelos, J. N. Fidalgo, and J. A. P. Lopes, A general approach
for security monitoring and preventive control of networks with large
wind power production, in Proc. 14th PSCC Power System Computation Conf., Sevilla, Spain, Jun. 2428, 2002, p. 1, Session 31, paper 2.
[19] V. Miranda, J. N. Fidalgo, J. A. P. Lopes, and L. B. Almeida, Real time
preventive actions for transient stability enhancement with a hybrid
neural networkOptimization approach, IEEE Trans. Power Syst.,
vol. 10, no. 2, pp. 10291035, May 1995.
[20] M. R. Aghamohammadi and A. Beik-Khormizi, Small signal stability
constrained rescheduling using sensitivities analysis by neural network
as a preventive tool, in Proc. IEEE PES Transmission and Distribution
Conf. Expo., New Orleans, LA, USA, Apr. 1922, 2010, pp. 15.
[21] A. D. P. Lotufo, M. L. M. Lopes, and C. R. Minussi, Sensitivity analysis by neural networks applied to power systems transient stability,
Elect. Power Syst. Res., vol. 77, pp. 730738, Jul. 2006, Elsevier Science.
[22] IEEE Committee Report, A Tutorial Course on Artificial Neural Networks With Applications to Power Systems. Piscataway, NJ, USA,
IEEE Press, 1996, 96TP112-0.
[23] Y. Mansour, E. Vaahedi, and M. A. El-Sharkawi, Dynamic security
contingency screening and ranking using neural networks, IEEE
Trans. Neural Netw., vol. 8, no. 4, pp. 942950, Jul. 1997.
[24] L. S. Moulin, A. P. A. da Silva, M. A. El-Sharkawi, and R. J. Marks, II,
Support vector machines for transient stability analysis of large-scale
power systems, IEEE Trans. Power Syst., vol. 19, no. 2, pp. 818825,
May 2004.
[25] K. Omata and K. Tanomura, Transient stability evaluation using an artificial neural network, in Proc. 2nd Int. Forum Applications of Neural
Networks to Power Systems, ANNPS93, 1993, pp. 130135.

2853

[26] A. W. N. Izzri, A. Mohamed, and A. Hussain, An improved method


in transient stability assessment of a power system using probabilistic
neural network, J. Appl. Sci. Res., vol. 3, no. 11, pp. 12671274, 2007,
INSInet Publication.
[27] R. Ebrahimpour and E. K. Abharian, An improved method in transient stability assessment of a power system using committee neural
networks, IJCSNS Int. J. Comput. Sci. Netw. Security, vol. 9, no. 1,
pp. 119124, Jan. 2009.
[28] R. Battiti, Using mutual information for selecting features in supervised neural net learning, IEEE Trans. Neural Netw., vol. 5, no. 4, pp.
537550, Jul. 1994.
[29] I. N. Kassabalidis, L. S. Moulin, M. A. El-Sharkawil, R. J. Marks, and
A. P. A. da Silva, Dynamic security border identification using enhanced particle swarm optimization, IEEE Trans. Power Syst., vol.
17, no. 3, pp. 723729, Aug. 2002.

Mauricio C. Passaro (M13) was born in So Paulo,


Brazil, on April 25, 1968. He received, in Brazil, the
B.Sc. and M.Sc. degrees from the Federal University of Itajub (UNIFEI) in 1994 and 2002, respectively, and the D.Sc. degree from Federal University
of Rio de Janeiro (UFRJ) in 2013, all in electrical
engineering.
He has worked in Alstom and VA Tech. He is currently with the Brazilian Independent System Operator (ONS) as a Senior Engineer. His area of expertise
is power systems dynamics and computational development of advanced power systems security applications.
Dr. Passaro is a Registered Professional Engineer in Brazil (CREA).

Alexandre P. Alves da Silva (M92SM00)


received the B.Sc. and M.Sc. degrees in electrical
engineering (EE) from the Catholic University of Rio
de Janeiro, PUC-Rio in 1984 and 1987, respectively,
and the Ph.D. degree in EE from the University of
Waterloo, Waterloo, ON, Canada, in 1992.
He was a Visiting Scholar at the University of
Washington, Seattle, WA, USA, in 1999. He worked
at the Federal University of Itajub from 1993 to
2002, and as a Full Professor at the Federal University of Rio de Janeiro (COPPE) from 2002 to 2011,
where he was Chairman of the Electrical Engineering Graduate Program. He
has published over 200 full papers in journals and conference proceedings.
He has worked regularly as a consultant and manager of research projects for
major companies in the Brazilian energy sector. Currently, he is the Leader of
the Center of Excellence in Smart Systems at the General Electrics Global
Research, Brazil Technology Center, in Rio de Janeiro.
Dr. Alves da Silva was the first Editor-in-Chief of the Brazilian Computational Intelligence Society Journal and he is Chairman of the Intelligent Systems
Subcommittee (PSACE) of the IEEE Power and Energy Society.

Antonio C. S. Lima (S95M00) was born in Rio


de Janeiro, Brazil, in 1971. He received the B.Sc.,
M.Sc., and D.Sc. degrees in electrical engineering
from the Federal University of Rio de Janeiro
(UFRJ) in 1995, 1997, and 1999, respectively.
In 1998, he was a Visiting Scholar with the
Department of Electrical and Computer Engineering
at the University of British Columbia, Vancouver,
BC, Canada. From 2000 to 2002, he was with the
Brazilian Independent System Operator, ONS, Rio
de Janeiro, Brazil, dealing with electromagnetic
transient studies for the Brazilian National Grid. Currently, he is an Associate
Professor with the Electrical Engineering Department, UFRJ, where he has
been since 2002.

You might also like