APPLICATION OF ARTIFICIAL INTELLIGENCE IN POWER TRANSFORMER FAULT DIAGNOSIS By YANG QI-PING , LI MENG-QUN, MU XUE-YUN, WANG JUN Presented by ARATHI AJAY S7 EEE NO: 7
APPLICATION OF ARTIFICIAL INTELLIGENCE IN POWER
TRANSFORMER FAULT DIAGNOSIS
ByYANG QI-PING , LI MENG-QUN, MU XUE-YUN, WANG JUN
Presented byARATHI AJAY
S7 EEENO: 7
CONTENTS
Artificial Intelligence Artificial Neural Network(ANN) Transformer fault and Dissolved Gas Analysis Transformer Fault Diagnosis Artificial Intelligence
(TFDAI) System TFD Expert System TFDANN Illustration Conclusion Reference
WHAT IS ARTIFICIAL INTELLIGENCE?
Study and design of intelligent systems. Intelligent system – one that perceives its environment
and takes actions to maximize its chances of success. Deals with reasoning, knowledge, planning, learning,
communication etc.
ARTIFICIAL NEURAL NETWORK
• AI implemented using ANN• Emulation of biological neural system• Imitates information processing of brain• Can teach oneself and adapt non-linear relations between
input and output.
NEURAL NETWORK
Biological Artificial
BIOLOGICAL NEURON
ARTIFICIAL NEURON
1)Dendrites 2)Electrical
impulse 3)Strength of
synapse 4)Cell body 5)Axon 6)Signal of axon
Input unit Input (p) Weight (W) Summation unit and transfer function Output unit Output (a)
TRANSFORMER FAULTS
Main faults areArcing or high current breakdown – H2, C2H2Low energy sparking or partial discharges –H2, other
lower HCHot spots or localized overheating – CH4, C2H6General overheating – CO, CO2
DISSOLVED GAS ANALYSIS
To detect incipient faults in oil-filled transformers Rate of evolution of HC temperature Methods: -
1. Key Gas Method
H2- Corona, O2 & N2- no fault, CO & CO2- cellulose breakdown
CH4 & C2H6- low temp oil BD, C2H4 – High temp BD, C2H2- Arcing
2. Ratio Method
CH4/H2, C2H2/C2H6, C2H2/C2H4
TRANSFORMER FAULT DIAGNOSIS ARTIFICIAL INTELLIGENCE SYSTEM
TFD Expert System TFD ANN
TFD EXPERT SYSTEM
Knowledge Base Database Inference Engine Interpretation Mechanism Man – Machine Interface
KNOWLEDGE BASE
Six Modules: -
Gas chromatography analysis moduleNormal, Normally aged, Partially discharging,
over heated – H2,CH4,C2H6,C2H4,C12, CO,CO2
Exterior Inspection moduleExterior or Interior Imperfection – noise, oil level, oil
temperature
Oil feature test moduleGood, alert, bad – acidity, resistivity, water content,
surface tension, dielectric loss, breakdown voltage
Insulation preventive test module
Insulation Status - Measure values of DC resistance, insulation resistance, leakage currents of high, medium, low voltage of 3phase winding
Comprehensive analysis module
Gives final judgment on analyzing the above modules
Coordinator moduleCoordinates and controls TFDES. Starts with
chromatography, then exterior inspection followed by oil feature and insulation preventive test modules.
TFD ARTIFICIAL NEURAL NETWORK
Two phases : 1. Learning2. Diagnosis
Five modules: 1.Characteristic Gas Method Module BP12.Three Ratios Method Module BP23.Insulation Oil Feature Test Module BP34.Exterior Inspection Module BP45.Comprehensive Analysis Module BP5
BP1 – 6inputs, 4outputs
I/p: H2, CH4, C2H2, C2H4, C2H6, CO
O/p: Normal, over heating, corona, arcing
BP2 – 3i/p, 9 o/p
I/p: C2H2/C2H4, CH4/H2,
C2H4/C2H6
O/p: 1 normal and 8 Faults
BP3 – 6 i/p, 3 o/p
I/p: acidity, resistivity, water content, surface tension, dielectric loss, breakdown voltage
O/p: good, alert, bad
BP4 – 3 i/p, 2 o/p
I/p: noise, oil level, oil temperature
O/p: exterior imperfection, interior imperfection
BP5 –
From the results of BP1, BP2, BP3, BP4, makes final judgment
TFDAI
Input - transformer’s measured data Preliminary judgment by logical judging model Status - either normal or abnormal If abnormal, TFDES and TFDANN starts parallel Their output to Comprehensive Analysis to give final
judgment
EXAMPLETest Data of a transformer:
CO
1264
Acidity .195
Resistivity 15x1010
Water content 33.5
Surface tension 20x10-3
Voltage rank 110
Breakdown voltage 39.5
Dielectric loss 1.75
From TFDES:
Transformer Interior Abnormal
Nature of fault: High Energy Discharge
Suggestion: stop operation immediately, make interior inspection
From TFDANN:
BP1 output: Y4=1.000, conclusion – electric arc discharge
BP2 output: Y4=0.998, conclusion – high energy discharge
BP3 output: Y1=.9895, conclusion – oil can be still used
BP4 output: Y1=.9385, conclusion – interior abnormal From TFDAI:
Transformer interior abnormal
Nature – high energy discharge
Conclusion - Transformer interior winding not good
Suggestion – stop operation, make interior inspection
On site conclusion – Winding Fault
CONCLUSION ES – to imitate logical thinking of brain ANN – to imitate thinking in images of brain DGA – powerful tool in fault diagnosis
*advantage
*disadvantageo TFDAI – remedy to the disadvantageso Results – highly reliable, less training time, less
memory consumption, 80% success levels
REFERENCES
Mu Xue-Yun, Wang Jun and Yang Qi-Ping, Li Meng-Qun. “Application of AI in Power Transformer Fault Diagnosis”, International Conference on Artificial Intelligence and Computational Intelligence, 2009, 7-8 Nov, p442-445
Martin T Hagan, Howard B Dcmuth and Mark Beale, “Neural Network Design”, chapter-2,3, page-2.1-2.23 and 3.1-3.16
Deepika Bhalla, RajKumar Bansal and Hari Om Gupta, “Appication of Artificial Intelligence Techniques for Dissolced Gas Analysis of Transformers- A Review”, World Academy of Science, Engineering and Technology 62 2010.
Methods and applications of Artificial Intelligence: vol3, pg 421-424, George A Vouros
K V Satyanarayana, C H Charkradhar Reddy, T P Govindan, Manoj Mandlik and T S Ramu, “Application of Artificial Intelligence for the assessment of the Status of Power Transformers”, Indian Institute of Science, Downloaded on March 9, 2010 from IEEE Xplore.
Jessey G Smith, B Venkba Rao, Vivek Diwanji and Shivaram Kamat, “Fault Diagnosis- Isolation of Malfunctions in Power Transformers”, june 10, 2009, Tata Consultancy Sevices. Copyright © 2009 Tata Consultancy Services.
THANK YOU