Sunteți pe pagina 1din 13

J.Sc.

Tech 12(1) 2011


159
Breast Cancer Diagnosis Using Artificial Intelligence Neural Networks
Howida Ali Abd Elgader
1
and Mohammed Hassen Hamza
2
1
Sudan University of Science and Technology
2
College of Computer Science & Information Technology
ABSTRACT: Breast cancer is the second largest cause of cancer deaths among women. Breast
cancer is one of the major tumor related cause of death in women. Various artificial intelligence
techniques have been used to improve the diagnoses procedures. The present study aimed at
detecting breast cancer by using three neural network topologies which are Multi Layer
Perceptron (MLP), Generalized Regression (GRNN) and Probabilistic (PNN). Three Wisconsin
breast cancer data sets (WBCD) were examined by these networks. It was seen that the most
suitable neural network model for classifying WBCD are MLP and GRNN, and observing that
PNN performance was low in case of small datasets.
KEYWORDS: Artificial neural networks, breast cancer, MLP, GRNN, PNN
INTRODUCTION:
It is now a well established fact that early
detection of cancer can play a significant role
in its treatment, making possible improvement
in the quality of the patients life and increase
survival rates. Early diagnosed, breast cancer
is one of the most treatable forms of cancer.
Most breast cancers are detected as a
lump/mass on the breast, or through self
examination or mammography [1].The
automatic diagnosis of breast cancer is an
important, real-world medical problem. A
major class of problems in medical science
involves the diagnosis of disease, based upon
various tests performed upon the patient.
When several tests are involved, the ultimate
diagnosis may be difficult to obtain, even for a
medical expert. This has given rise, over the
past few decades, to computerized diagnostic
tools, intended to aid physicians in making
sense out of the confusion of data [1].Various
artificial intelligence techniques have been
used to improve the diagnoses procedures and
to aid the physicians efforts [2] .Artificial
Neural Networks are systems inspired by the
human brain. They are densely interconnected
networks of Processing Elements together
with rule to adjust the strength of the
connections between the units in response to
externally supplied data [11].
Medical applications of ANNs are mostly
based on their ability to handle classification
problems including classifications of illnesses
or to estimate prognosis [3].Neural networks
are highly adaptive structures with numerous
adaptable coefficients which have to be set on
specific values. This adaptation is called a
training process. The training process is
applied to train the network offline to perform
specific mapping that is required. During the
training process a representative set of data is
presented to the network. The training process
is finished when a preset network mapping
accuracy is reached. Most types of networks
need exact correct output network value for
each input value during training process [4].It
means that the output data set has to be given
as a desired output value for each input data.
That, way for one set of input data the
network may be trained to perform arbitrary
input-output mapping. After the trained
process is finished, network input-output
mapping is memorized and network may be
tested with a test data set. Test data sets
J.Sc. Tech 12(1) 2011
160
usually have different values from training
data sets. That happens when noise is
presented in the data set or when training and
test data sets are defined on continuous
domains [4]. Since the network is trained on a
specific training data set for arbitrary new (not
memorized) inputs, network outputs are
closest to memorized output. In other words,
networks have filtering properties in presence
of noise and property of input data
classification into predefined subsets. This is
an important network property which is
commonly used. After the training process is
finished, the network is ready to perform
required mapping with predefined accuracy.
Neural Network Perceptron Layer
Network : Perceptron layer network is one of
the models in the design of neural networks.
The model contains an input layer, a hidden
layer and an output layer, where the hidden
layer is optional [4]. The input layer receives
the input. These inputs are multiplied by
appropriate weights [5], and the obtained
value is checked against a large set of data.
The weights are adjusted and a new value is
calculated. This step is repeated until the
desired results are obtained [6]. The neural
network structure most commonly used for
classification tasks, is Multi-Layer
Perceptrons.
Back Propagation Learning Algorithm :
The back propagation algorithm trains a given
feed-forward multilayer neural network for a
given set of input patterns with known
classifications. When each entry of the sample
set is presented to the network, the network
examines its output response to the sample
input pattern. The output response is then
compared to the known and desired output and
the error value is calculated. Based on the
error, the connection weights are adjusted.
The back propagation algorithm is based on
Widrow-Hoff delta learning rule in which the
weight adjustment is done through mean
square error of the output response to the
sample input. The set of these sample patterns
is repeatedly presented to the network until the
error value is minimized [7]. The technique is
summarized as follows :
1.Present a training sample to the neural
network.
2.Compare the network's output to the desired
output from that sample, calculate the error in
each output neuron.
3.For each neuron, calculate what the output
should have been, and a scaling factor, how
much lower or higher the output must be
adjusted to match the desired output. This is
the local error.
4.Adjust the weights of each neuron to lower
the local error.
5.Assign "blame" for the local error to neurons
at the previous level, giving greater
responsibility to neurons connected by
stronger weights.
6.Repeat the steps above on the neurons at the
previous level, using each one's "blame" as
its error.
Actual algorithm:
Initialize the weights in the network (often
randomly)
Do
For each example e in the training set
O = neural-net-output(network, e) ; forward
pass
T = teacher output for e
Calculate error (T - O) at the output units
Compute delta_wi for all weights from
hidden layer to output layer ; backward pass
Compute delta_wi for all weights from input
layer to hidden layer ; backward pass
continued
Update the weights in the network
Until all examples classified correctly or
stopping criterion satisfied . Return the
network as the algorithm's name implies, the
errors (and therefore the learning) propagate
backwards from the output nodes to the
inner nodes. So technically speaking, back
propagation is used to calculate the gradient
J.Sc. Tech 12(1) 2011
161
of the error of the network with respect to
the network's modifiable weights. This
gradient is almost always used in a simple
stochastic gradient descent algorithm to find
weights that minimize the error. The term
"back propagation" is often used in a more
general sense, to refer to the entire
procedure encompassing both the
calculation of the gradient and its use in
stochastic gradient descent. Back
propagation usually allows quick
convergence on satisfactory local minima
for error in the kind of networks to which it
is suited [8].
The Generalized Regression Neural
Networks (GRNN)
The GRNN consists of four layers: input
layer, pattern layer, summation layer, and
output layer. The number of input units in
the first layer is equal to the total number of
parameters. The first layer is fully connected
to the second, pattern layer, where each unit
represents a training pattern and its output is
a measure of the distance of the input from
the stored patterns. Each pattern layer unit is
connected to the two neurons in the
summation layer: S-summation neuron and
D-summation neuron. The summation
neuron computes the sum of the weighted
outputs of the pattern layer while the D-
summation neuron calculates the un
weighted outputs of the pattern neurons. The
connection weight between the ith neuron in
the pattern layer and the summation neuron
is y
i
; the target output value corresponding
to the ith input pattern. For D-summation
neuron, the connection weight is unity. The
output layer merely divides the output of
each S-summation neuron by that of each D-
summation neuron, yielding the predicted
value to an unknown input vector [9].
Probabilistic Neural Network
The probabilistic neural network is a direct
continuation of the work on Bayes
classifiers. More precisely, the PNN is
interpreted as a function which
approximates the probability density of the
underlying examples distribution (rather
than the examples directly by fitting) [10].
The PNN consists of nodes allocated in
three layers after the inputs:
- pattern layer: there is one pattern node for
each training example. Each pattern node
forms a product of the weight vector and the
given example for classification, where the
weights entering a node are from a particular
example. After that, the product is passed
through the activation function:
ct| (
T

1 ) / o
2
|
- summation layer: each summation node
receives the outputs from pattern nodes
associated with a given class:
L

ct| (
T

1 ) / o
2
|
- output layer: the output nodes are binary
neurons that produce the classification
decision
L

ct| (
T

1 ) / o
2
| > L

ct|
(
T

1 ) / o
2
|
The only factor that needs to be selected for
training is the smoothing factor, that is the
deviation of the Gaussian functions:
- too small deviations cause a very spiky
approximation which can not generalize
well;
- too large deviations smooth out details.
An appropriate deviation is chosen by
experiment [10]
MATERIAL
The material that was used in the present
work was derived from the internet site of
University of California at Irvine (UCI)
Machine Learning Data Repository
3
. The
downloaded file contains medical data
J.Sc. Tech 12(1) 2011
162
concerning breast cancer classification cases
that were categorized by medical experts to
malignant or benign. Table (1) show the
discretion of the three data sets:
Table 1. Data discretions
Data Set Number Number of Instance Number of
Attributes
Data Type Class Distribution
1 699
Missing values= 16
10 Integer Benign:458
Maliganant:241
2 569 32 Real Benign:357
Maliganant:212
3 198 35 Real Benign:151
Maliganant:47
Table 2. Classes and their data discretions
Class B
test
Class A
test
Class
B
Train
Class A
train
Train data/Test data Dataset Number
73 133 166 311 477/206 1
64 107 148 250 398/171 2
44 14 104 32 136/58 3
SIMULATION RESULTS :
One of the most difficult problems for
neural network modeling is selection of
proper neural network structure. Three
neural network structures, multi-layer
perceptrons (MLP) with three layers and 15
hidden nodes, generalized regression neural
networks (GRNN) and probabilistic neural
network (PNN) were applied to three
WBCD data sets to show the performance of
neural networks on breast cancer data. The
simulations were developed using
MATLAB 7.0 Neural Network Toolbox.
The results are shown in table .3.
Table 3. ANNs Performance
Data Set MLP GRNN% PNN%
1 99 97 99
2 98 95 96
3 70 75 75
CONCLUSION :
The neural networks have been applied with
successfully for pattern classification and
recognition problems . The idea as a whole
are a decision making problem, in which a
group of previous observations for the
patient state variable are used to identify the
class of breast cancer. How the neural
networks are used in actual clinical
diagnosis of breast cancer is shown in this
paper. The performance of the three
networks MLP,GRNN and PNN was
investigated for breast cancer diagnosis
using three data sets. MLP, GRNN and PNN
are best classifiers for all data set. The third
data set gave different result because
number of instance was less than the other
sets. According to overall results, it was
seen that the most suitable neural network
model for classifying WBCD data are MLP
and GRNN. The work also indicated that
neural networks can be effectively used for
breast cancer diagnosis to help oncologists
REFERENCES:
1.Tba Kiyan, Tlay Yildirlm (2004).
Breast cancer diagnosis using statistical
neural networks, Journal of Electrical and
Electronics Engineering,14, 1149-1153.
J.Sc. Tech 12(1) 2011
163
2.Spline- B . ,Yuehui Chen, Mingjun Liu,
and Yang , B. O. (2006), Breast cancer
detection using hierarchical, ICNC, Part I,
LNCS 4221, pp. 2528,.
3.Truong, H. ( 2007), Neural networks as
an aid in the diagnosis of lymphocyte-rich
effusions.
4.Markley, Tourassi, M. L. J. and Floyd,
G. C. (2003). Self organizing map for cluster
analysis of a breast cancer database, Artif
Intell Med, pp. 113-127.
5.Kohonen, T., Kangas, J. A., Laaksonen,
J.T.(1990).Variants of self-organizing
maps, IEEE Trans on Nerual Networks,
I,93-99.
6.Pradeep. N. ,Shankaragowda.B. B.,
Karibasappa .K and Ramaswamy. V., An
Artificial Neural Network Approach for
mammography images feature extraction,
Proceedings of the international Conference
on Cognition and Recognition, pp.768-773.
7.Bose, N. K. and Liang, P.(1996). neural
network Fundamentals with Graphs,
Algorithms, and Applications, McGraw-
Hill, New York, NY.
8.Duda , R. O. , Hart , and Stork , P. (2000)
Pattern Classification, 2ed., Wiley .
9.Kerem .H CIIZOLU1 Pnar
AKIN(1990) Artificial neural network
models in : Rain fall Run - off Modeling of
Turkish Rivers.
10. Specht, D. F. P. Robabilistic Neural
Networks, Neural Networks, vol. 3, pp.109-
118 .
11. Soman, K. P. , Shyam Diwakar and
Ajay. V. (2006), Data Mining Theory and
Practice, Prentice Hall of India .
J.Sc. Tech 12(1) 2011
164
Design of Fuzzy Power System Stabilizer for Single Machine System
Eltahir Mohamed Hussin
1
and Bahar A. A. Elmahi
2
1
Sudan University of Science and Technology, College of Engineering, Biomedical Engineering Department
Khartoum Sudan.
2
Industrial Research and Consultancy Center Khartoum Sudan.
ABSTRACT: This paper presents a real- time fuzzy logic controller capable of
improving power system stability. This is done by means of the fuzzy membership
functions whose values are computed depending on the measured generator frequency
and voltage. The controller was tested in real-time for different loads on the generator.
Simulation results showed that the designed fuzzy controller realized a good
dynamic behavior of the system, no overshoot and a good rejection of impact loads
disturbance. The results gave better performance and higher robustness than those
obtained by the application of a conventional controller.
KEYWORDS: Synchronous machine, Power system stability, fuzzy logic controller.
.
INTRODUCTION
Traditionally, controllers are designed
on the basis of a mathematical
description of a system and its
linearized model. It is well known that
power system stabilizers (PSSs) on
generation units are effective tools to
damp electromechanical oscillation
and enhance transient stability in
electric power systems. However, the
long-term operating experience has
shown that conventional PSSs do not
always completely use the capabilities
of excitation system, because
parameters of controllers are fixed and
tuned only for a specific operation
point. Conventional control may be
modified based on results of testing
and experience, but due to the
difficulties of analysis many such
controllers ended in being linear
[1]
.
Recently, new approaches to designing
PSS's have been proposed as a result of
developing computer technologies
including Artificial Intelligence (AI)
technique such as fuzzy logic, neural
network . [
2,3
]. The most recent
progress in field of AI application has
been made in the development of
Fuzzy logic power system stabilizer
(FLPSS) . Fuzzy logic control has
emerged as a powerful tool and is
starting to be used in various power
system applications
[1-8]
. The
application of fuzzy logic control
technique appears to be most suitable
one. whenever a well-defined control
objective cannot be specified the
system to be controlled is a complex
one, or its exact mathematical model is
not available. Hence fuzzy controller
uses fuzzy logic as a design
methodology which can be applied in
developing linear and nonlinear
systems for embedded control. It has
been found to be a good replacement
for conventional control techniques
which require highly complicated
mathematical models. However, the
method simplicity enables the control
designers to realize a control in less
development time at lower cost, and
better performance.
This paper presents adaptive a fuzzy
controller as power system stabilizer
for different operating conditions of
the power system. Fuzzy logic
approach, for system identification
using input and output data (all data in
appendix A) and its application to
design a FPSS. The simulation starts
with output changeable loop of the
conventional controller of the system
J.Sc. Tech 12(1) 2011
165
in order to compare the performances
to those obtained from the respective
FLC system. The system is modeled
and simulated using TIl Shell software.
The system has been performed in
order to subject it to several types of
large disturbances using a single-
machine infinite bus power system.
The performance of the system with
FPSS is compared with the system
having conventional PSS.
System Model: The system
configuration for single machine
infinite bus system considered is
shown in Fig.1. The system studied
includes a single synchronous
generator, automatic voltage regulator,
and governor control. A
stabilization problem is to design a
power system stabilizer that provides a
supplementary stabilizing signal to
enhance the damping torque of the
power system. The design
methodology depends on forming
general rule membership functions and
on determining parameters based on
observed response of system to
disturbance.
Fig. 1 Single Machine Infinite Bus
System configuration
FLC structure: In the conventional
control, the amount of control is
determined in relation to a number of
data inputs using a set of equations to
express the entire control process.
Expressing human experience in the
form of a mathematical formula is a
very difficult task, if not impossible.
Fuzzy logic is derived from fuzzy set
theory where infinite number of
memberships is allowed, it provides a
simple tool to interpret this experience
into reality. Fuzzy logic controllers are
rule-based controllers. The structure of
the FLC resembles that of a knowledge
based controller except that the FLC
utilizes the principles of the fuzzy set
theory in its data representation and its
logic. The basic configuration of the
FLC can be simply represented in;
fuzzification, knowledge base (Rule,
decision) and defuzzufication module
which converts the inferred decision
from the linguistic variables back to
the numerical values Fig.2. The basic
concepts of fuzzy logic control theory
were described by Hussien et.al
[10] and carcau [11]. The operation
principle of such devices is well-
known and described in many
publications. Figure 2 shows Fuzzy
Logic PSS structure.
Fuzzy controller design: The control
design paradigm is to form a system
model and develop control laws from
analysis of the model. The controller
may be modified based on results of
testing and experience. The precise
fuzzy membership functions (MFs)
depend on the valid range of inputs and
general response of the system
[6,10,11]. The design process of FLC
consists of the following steps:-
Fig .2 Fuzzy Logic PSS structure
a - Identification of I/O variables
(Fig.3).
b - Choose the membership function.
J.Sc. Tech 12(1) 2011
166
Fuzzy sets are defined for each input
and output variable. There are three
fuzzy levels
Fig 3. Designing I/O on fuzzy system
/TILShell software.
for inputs and two for output. The
membership functions for input and
output variables are triangular. The
min - max method inference engine is
used; in practice, the universe of
discourse is restricted to a
comparatively small interval [Xmin,
Xmax]. The universe of discourse of
each fuzzy variable can be quantized
into a number of overlapping fuzzy
sets (linguistic variables). The number
of fuzzy sets for each fuzzy variable
varies according to the application.
The number of fuzzy sets results in a
corresponding increase in the number
of rules
[12]
. According to the Fig. 4
showing the fuzzy membership
functions; three linguistic variables for
fuzzification and two for
defuzzification. The whole changing
(V,Hz) space is mapped with decision
variables obtained form expert
information. This decision table
considers all possible control values.
Hence the fuzzy relationship matrix is
constructed out of the 9 possible input
combinations called states, against two
possible decisions in a 9x2 matrix.
c Rules Base
The main part of the FLC is the Rule
Base and the Inference Mechanism.
Here, uncertainties are associated with
each rule in the rule-base. The rule
base is normally expressed in a set of
Fuzzy Linguistic rules, with each rule
triggered with varying
Fig 4. MFs inputs , voltage (a) ,
frequency (b) and output (c)
belief for support. The control rules
were designed to be symmetric under
the assumption that if necessary any
asymmetries could be best handled
through scaling Fig.4 [12,13].
Defuzzification strategy
Defuzzification is a process of
converting the FLC statement into
control actions, or from fuzzy values to
crisp values. The performance of the
FLC depends very much on the
deffuzzification process.
Defuzzification interface performs the
following functions: a) It executes
defuzzification. b) It transforms output
variable value into the actual control-
input of the universe of discourse[17].
The method used here is the center of
area (COA) method. It is also called
the center of gravity method.
a
b
c
J.Sc. Tech 12(1) 2011
167
As mentioned in [12][13]
.
Various
defuzzification methods have been
proposed to convert the output of the
fuzzy controller to a crisp value
required by the plant. These methods
are: Center of Area (COA), Center of
Sum (COS), Height Method (HM),
Mean of maxima (MOM), Center of
Largest (COLA) and First maxima
(FM).
Fig 5. Tuning rule
Fuzzy logic based PSS: The power
system stabilizer is used to improve the
performance of synchronous
generators. However, poor
performance under various loading
conditions is obtained with
conventional PSS. Therefore, the need
for fuzzy logic PSS arises. The fuzzy
controller used in power system
stabilizers is normally a two-input and
a single-output component. A
modification of feedback voltage to
excitation system and frequency as a
function of accelerating power on a
unit is used to enhance the stability of
the system. The stabilizing signals are
computed using the standard fuzzy
membership functions depending upon
these variables[15]. A particular
control design requires specification of
all control rules and membership
functions. The control rules are based
on an understanding of the desired
effect of the controller.
The traditional PSS is modeled by the
transfer function of Eq.1 [14]. The first
term in Eq.1, is a reset term that is used
to wash out the compensation effect
after a time lag T. The second term of
Eq.1 is a lead compensator to account
for the phase lag through the electrical
system [14]. In many practical cases,
the phase lead required is greater than
that obtainable from a single lead
network.
In this case, cascaded lead stages are
used where k is the number of lead
stages. For the proposed FPSS the
second term of eq.1 is replaced with a
fuzzy logic rule-base using the
deviation from synchronous voltage
and frequency deviation of the
machine are the error, e, and error
change signals, respectively for the
controller [14-19]. The control output
u, is the stabilizing signal Vs. Each
control rule R, is of the form:
IF e is A, AND is B, THEN u is C
(2)
Simulation Results and Discussion:
The proposed approach applied to
single machine infinte bus system(Gas
Turbine) is one of system existing in
Gas Plant Station-Garri area. In this
paper ,the simulation was implemented
by using TIL Shel software in fuzzy
system environment. The data of the
system are given in the appendix A.
The system was simulated with
different values of the parameter
considered and compared to nominal
value (real value).The feedback control
algorithms were iterated until best
(1)
J.Sc. Tech 12(1) 2011
168
simulation results were obtained
(Fig.6,7 and 8).
Fig.9 shows that the FPSS has an
excellent response to different load
disturbance, compared to CPSS
response. Table .1; shows the outputs
for both controllers (CPP,FLC). The
reason for superior performance of
fuzzy controlled system is that
basically it is adaptive in nature and
the controller is able to realize
different control law for each input
state.
Table 1. The simulation results
Time Conventional FLC Time Conventional FLC
1:04 20.92 20.6917 1:34 20.62 20.6917
1:06 20.62 20.6917 1:36 20.62 20.6917
1:08 20.92 20.6917 1:38 21.22 20.6917
1:10 20.62 20.6917 1:40 20.77 20.6917
1:12 21.12 20.6917 1:42 20.62 20.6917
1:14 21.12 20.6917 1:44 20.77 20.6917
1:16 20.82 20.6917 1:46 20.92 20.6917
1:18 20.62 20.6917 1:48 20.62 20.6917
1:20 21.12 20.6917 1:50 20.62 20.6917
1:22 21.07 20.6917 1:52 20.62 20.6917
1:24 20.92 20.6917 1:54 21.22 20.6917
1:26 20.92 20.6917 1:56 20.92 20.6917
1:28 20.62 20.6917 1:58 20.62 20.6917
Fig 6. The performance of CPSS
J.Sc. Tech 12(1) 2011
169
Fig 7. The performance of FPSS
Fig.8 The different between two methods (CPSS, FPSS)
Fig. (9) Comparing the two methods (FPSS&PSS)
Fig 10. Control Surface rule viewer
CONCLUSION
This work studied the effect of the
parameters uncertainties on the power
system stability.The control strategy is
based on FLC approach. A complete
fuzzy logic control based synchronous
generator has been described. The
system was analyzed and designed,
and performance was studied
extensively by simulation in order to
validate the theoretical concept. The
simulation results show that the
proposed controller is superior to
conventional controller in robustness
and in tracking precision. The
simulation study clearly indicates the
superior performance of fuzzy control,
because it is inherently adaptive in
nature. It appears from the response
properties that it has a high
performance in presence of the plant
parameters uncertain and load
disturbances. FLC gives fast dynamic
response with no overshoot and
negligible steady-state error. The
CPSS FPSS
J.Sc. Tech 12(1) 2011
170
decoupling, stability and convergence
to equilibrium point are verified.
APPENDIX A
Time Voltage Frequency Output
MW
Time Voltage Frequency Output
MW
1:14 10.45 50.0 21.12 1:30 10.45 50.00 21.22
1:16 10.45 50.0 20.82 1:32 10.45 50.00 21.22
1:18 10.45 50.0 20.62 1:34 10.45 50.01 20.62
1:20 10.45 50.0 21.12 1:36 10.45 50.00 20.62
1:22 10.45 50.0 21.07 1:38 10.45 50.01 21.22
1:24 10.45 50.0 20.92 1:40 10.45 50.00 20.77
1:26 10.45 50.0 20.92 1:42 10.52 50.00 20.62
1:28 10.52 50.0 20.62 1:44 10.52 50.00 20.77
REFERENCES:
1. Etingov. P. V. and Voropai, N. I.
(2006). Application of fuzzy logic
PSS to enhance transient stability
in large power systems. The
International Conference on Power
Electronics, Drives and Energy
Systems for Industrial Growth-,
New Delhi, India, 1215.
2. Hiyama, T. (1990). Rule-based
stabilizer for multi-machine power
system. IEEE Transactions on
Power Systems, pp.403409.
3. Zhang, Y. O. P. Malik and Chen,
G. P (1995). Artificial neural
network power system stabilizers
in multi- machine power system
environment. IEEE Transactions
on Energy Conversion., pp. 147-
153.
4. Chaturvedi. D. K, and Malik. O. P.
(2005). Generalized neuron-based
adaptive PSS for multi machine
environment. IEEE Transactions
on Power Systems, pp. 358-366.
5. Handschin. E. Hoffmann, W. and
Reyer, F. (1994). A new method
of excitation control based on
fuzzy set theory. IEEE
Transactions on Power Systems,
pp.533-539 .
6. Hiyama,T. (1999). Development
of fuzzy logic power system
stabilizer and further studies.
Proceedings of International
Conference on Systems, Man and
Cybernetics IEEE SMC'99, Tokyo,
Japan.
7. Sabapath,. D. (2008). Design of
fuzzy based power system stability
for a multi- machine system.
International Journal of Electric
and Power Engineering 248-253..
8. Abdel Ghani Aissaoui (2007) . A
fuzzy logic controller for
Synchronous Machine" Journal of
Electrical Engineering, pp285-290.
9. Neeraj Gupta, (2008). "Fuzzy logic
based power system stabilizer.
Patiala 147004 .
10. Hassan, M. A. M . , Malik , O. P.
and Hope, G. S. (1991). A Fuzzy
logic based stabilizer for a
Synchronous Machine. IEEE
Transactions on Energy
conversion, pp. 407-413,.
11. Jenana Ileana Corcau (2007).
Fuzzy logic controller as a power
system stability. International
Journal of Circuits System and
Signal Processing,
12. Tomescu, B. (2001). On the use of
fuzzy logic to control paralleled dc-
dc converters. Diss ertation
Virginia Polytechnic Institute and
State University Blacksburg,
Virginia.
13. Kundur, P. (1994). Power System
Stability and Control, McGraw-
Hill.
14. Hoang , P. and Tomsovic, K.
(1996). Design and analysis of an
adaptive fuzzy power system
J.Sc. Tech 12(1) 2011
171
stabilizer . IEEE Transactions on
Energy Conversion, Vol. 11, No. 2,
15. Zimmerman, H. J (1985). Fuzzy
set theory and its application.M C
graw Hill.
16. Hasan , A. and Sadrul Ula. A. H.
M. (1994). Design and
implementation of a fuzzy
controller based automatic voltage
regulator for a synchronous
generator. IEEE/ PES Winter
Meeting, 94 WM 025-7, New
York.
17. Noroozian, M , Andersson ,G. and
Tomsovic, K (1995). Robust,
Near time-optimal control of
power system oscillations with
fuzzy logic. IEEE@ES winter
Meeting, 95 WM237-8 PWRD.
New York.
18. Tomsovic, K . and Hoang, P.
(1994). Approaches for evaluating
fuzzy logic based power system
stabilization control, Rough Sets
and Soft Computing Svmposlum,
pp. 262-269.

S-ar putea să vă placă și