Escolar Documentos
Profissional Documentos
Cultura Documentos
I. INTRODUCTION
Manuscript received December 22, 2005; revised August 14, 2006. Paper no.
TPWRD-00724-2005.
The authors are with the Key Laboratory of High Voltage and Electrical
New Technology of Ministry of Education, Chongqing University, Chongqing
400044, China.
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/TPWRD.2007.893182
931
932
(3)
933
affinity is, the smaller the mutation rate will be. Equation
(4) is shown as follows:
(4)
and
4) Affinity calculation: Determine the affinity among
. From
, reselect % of the antiall the elements of
bodies with the highest affinity as the set of partial anti. At the same time, eliminate all of the memory
bodies
whose affinities are smaller than the
antibodies from
natural death threshold .
5) Clonal suppression: Eliminate those memory clones of the
, whose affinity
is superior to
same fault type from
the suppression threshold and generate the set of partial
. The total memory antibody pool
memory antibodies
can be obtained as follows:
.
6) Antigen presentation: Select the next antigen and return
to (1). This process continues until all antigens have been
presented to the network.
, eliminate those memory
7) Network suppression: From
is suantibodies of the same fault type whose affinity
perior to until the discrepancy of the average affinity is
inferior to a prescribed threshold in four consecutive iterations. If the stopping criterion has not been met, then turn
to (8).
8) Randomly generating antibody: Randomly generate anand add them to
, so antibodies set
is
tibodies
equal to
, then turn to (1) to repeat the entire
process.
C. Improved KNN Classifier
The KNN is a simple but effective classification method.
It assigns an input pattern to the majority class according to
nearest
Euclidean distances between the pattern and its
neighbors in the training data space. After fulfilling the artificial immune network training, the evolved memory antibodies
are available for KNN classification. Through calculating the
Euclidean distance between test antigen and memory antibodies, the fault type of the test antigen is determined by
most stimulated
using a majority vote of the outputs of the
memory antibodies. But that is only applicable to single fault
diagnosis. Since multiple faults antigens have almost equal
distances among several fault types of memory antibodies, only
the nearest memory antibody votes would result in an incorrect
diagnosis. After many tests, we can draw the conclusion that
the fault type of the test antigen is determined by both memory
antibodies if the discrepancy of the Euclidean distance between
test antigen and two nearest memory antibodies is inferior to
10% of their average distance. If the test antigen obtains the
same votes, multiple incipient faults are considered to occur
simultaneously.
IV. PARAMETERS ANALYSIS
In this section, we intend to discuss and analyze how sensitive
the algorithm is to some user-defined parameters. In particular,
we will study the influence of the parameters and to the convergence speed, final network size, and recognition accuracy.
934
TABLE I
COLLECTED FAULT SAMPLES OF VARIOUS TYPES
TABLE II
SOME FAULT MODES OF HIGH ENERGY DISCHARGE. (AFFINITY
0.8)
TABLE III
PROCESSING OF SINGLE-FAULT SAMPLES BASED ON AINC AND ITS
DIAGNOSIS ACCURACY FOR THE POWER TRANSFORMER
Fig. 4. Shape space of high energy discharge and its memory antibodies.
Fig. 5. Shape space of MF1 and its memory antibodies of T3 and D2.
TABLE IV
MULTIPLE INCIPIENT FAULTS DIAGNOSIS
ACCURACY OF THE POWER TRANSFORMER
TABLE V
FAULT DIAGNOSIS OF EIGHT REPRESENTATIVE SAMPLES USING AINC
935
[7] Q. Su, C. Mi, L. L. Lai, and P. Austin, A fuzzy dissolved gas analysis
method for the diagnosis of multiple incipient faults in a transformer,
IEEE Trans. Power Syst., vol. 15, no. 2, pp. 593598, May 2000.
[8] K. Spurgeon, W. H. Tang, Q. H. Wu, Z. Richardson, and G. Moss,
Dissolved gas analysis using evidential reasoning, Proc. Inst. Elect.
Eng., Sci. Meas. Technol, vol. 152, no. 3, pp. 110117, 2005.
[9] C. E. Lin, J. M. Ling, and C. L. Huang, An expert system for
transformer fault diagnosis using dissolved gas analysis, IEEE Trans.
Power Del., vol. 8, no. 1, pp. 231238, Jan. 1993.
[10] Z. Wang, Y. Liu, and P. J. Griffin, A combined ANN and expert
system tool for transformer fault diagnosis, IEEE Trans. Power Del.,
vol. 13, no. 4, pp. 12241229, Oct. 1998.
[11] H.-T. Yang and C.-C. Liao, Adaptive fuzzy diagnosis system for dissolved gas analysis of power transformers, IEEE Trans. Power Del.,
vol. 14, no. 4, pp. 13421350, Oct. 1999.
[12] A. S. Farag, M. Mohandes, and A. Al Shaikh, Diagnosing failed distribution transformers using neural networks, IEEE Trans. Power Del.,
vol. 16, no. 4, pp. 631636, Oct. 2001.
[13] Y. Zhang, X. Ding, Y. liu, and P. J. Griffin, An artificial neural
network approach to transformer fault diagnosis, IEEE Trans. Power
Del., vol. 11, no. 4, pp. 18361841, Oct. 1996.
[14] J. L. Guardadao, J. L. Naredo, P. Moreno, and C. R. Fuerte, A comparative study of neural network efficiency in power transformers diagnosis using dissolved gas analysis, IEEE Trans. Power Del., vol. 16,
no. 4, pp. 643647, Oct. 2001.
[15] K. F. Thang, R. K. Aggarwal, A. J. McGrail, and D. G. Esp, Analysis
of power transformer dissolved gas data using the self-organizing map,
IEEE Trans. Power Del., vol. 18, no. 4, pp. 12411248, Oct. 2003.
[16] H.-T. Yang, C.-C. Liao, and J.-H. Chou, Fuzzy learning vector quantization networks for power transformer condition assessment, IEEE
Trans. Dielect. Electr. Insul., vol. 8, no. 1, pp. 143149, Feb. 2001.
[17] Y. C. Huang, Condition assessment of power transformers using
genetic-based neural networks, Proc. Inst. Elect. Eng., Sci. Meas.
Technol., vol. 150, no. 1, pp. 1924, 2003.
[18] Y.-C. Huang, A new data mining approach to dissolved gas analysis
of oil-insulated power apparatus, IEEE Trans. Power Del., vol. 18, no.
4, pp. 12571261, Oct. 2003.
[19] Y.-C. Huang and C.-M. Huang, Evolving wavelet networks for power
transformer condition monitoring, IEEE Trans. Power Del., vol. 17,
no. 2, pp. 412416, Apr. 2002.
[20] V. Miranda, A. Rosa, and G. Castro, Improving the IEC table for transformer failure diagnosis with knowledge extraction from neural networks, IEEE Trans. Power Del., vol. 20, no. 4, pp. 25092516, Oct.
2005.
[21] W.-M. Lin, C.-H. Lin, and M.-X. Tasy, Transformer-fault diagnosis
by integrating field data and standard codes with training enhancible
adaptive probabilistic network, Proc. Inst. Elect. Eng., Gen. Transm.
Distrib., vol. 152, no. 3, pp. 335341, 2005.
[22] Z. Yongli, W. Lizeng, and L. Xueyu, Synthesized diagnosis on transformer faults based on Bayesian classifier and rough set, in Proc.
CSEE, 2005, vol. 25, no. 10, pp. 159165.
[23] L. N. de Castro and F. J. Von Zuben, An evolutionary immune system
network for data clustering, in Proc. 6th Brazilian Symp. Neural Networks, Rio de Janeiro, Brazil, 2000, pp. 8489.
[24] L. N. de Castro and F. J. Von Zuben, Learning and optimization using
the clonal selection principle, IEEE Trans. Evol. Comput., vol. 6, no.
3, pp. 239251, Jun. 2002.
[25] N. K. Jerne, Towards a network theory of the immune system, Annu.
Immunol. (Inst. Pasteur), vol. 125, no. C, pp. 373389, 1974.
Xiong Hao was born in Nanchang, China, on May 6, 1979. He received the M.S.
degree in power engineering from Chongqing University, Chongqing, China, in
2003, where he is currently pursuing the Ph.D. degree in the key lab of high
voltage and new electrical technology.
His research area is online monitoring and fault diagnosis for high-voltage
(HV) apparatus.