Diapositive 1

1
The Use of Artificial Neural Networks for
Identification and Location of Defective
Insulators in Power Lines Through Current
Transformers
Renato Massoni Capelini UNIFEI Brazil Manuel
Luiz Barreira Martinez UNIFEI - Brazil José
Feliciano Adami UNESP Brazil Marcel Fernando
da Costa Parentoni UNIFEI - Brazil Ithamar Sene
Junior AES Sul - Brazil
CAPELINI BR RIFS3 0572
2
Purpose of the Project

• Remote identification and location of defective
  insulator strings
• Use the current transformer as a capacitive
  coupler for signal acquisition

CAPELINI BR RIFS3 0572
3
Defective Insulator Strings
CAPELINI BR RIFS3 0572
4
Corona Effect

• Partial Discharge
• Low Amplitude
• High Frequency

CAPELINI BR RIFS3 0572
5
Corona Effect Emulation
CAPELINI BR RIFS3 0572
6
The Laboratory
48 km Artificial Transmission Line 48 Taps in the
Transmission Line (4 taps/pi)
CAPELINI BR RIFS3 0572
7
The Laboratory
CAPELINI BR RIFS3 0572
8
CT as a Coupling Device
CAPELINI BR RIFS3 0572
9
CT as a Coupling Device
CAPELINI BR RIFS3 0572
10
AI Neural Networks

• Based on the Human Brain
• Good for Pattern Recognition

CAPELINI BR RIFS3 0572
11
Software Development
CAPELINI BR RIFS3 0572
12
Software Development Test 1

• Location of the Damaged String in the Power Line
• String 1 Defective Insulator
• 12 Taps of the Power Line (Step 4 km)
• Data Bank 12000 Signals from the Damaged String

CAPELINI BR RIFS3 0572
13
Software Development Test 1
Tap 1
Tap 5
Tap 9
Tap 13
Tap 17
Tap 45
CAPELINI BR RIFS3 0572
14
Software Development Test 1
CAPELINI BR RIFS3 0572
15
Conclusion

• CT as a Coupling Device ? Saving
• Neural Network ? Very Powerful and Precise Tool
• Data Bank Size ? Direct Influence on the Results

CAPELINI BR RIFS3 0572
16
Thank You!
CAPELINI BR RIFS3 0572