Barcelona 12 15 May 2003 Kerman Regional Electric
Barcelona 12 -15 May 2003 Kerman Regional Electric Company & Fault Location in Distribution Systems based on Artificial Neural Networks and Application of GIS M. Zangiabadi University of Tehran Co. Mansooreh Zangiabadi M. R. Haghifam A. Khanbanha Tarbiat Modares University Kerman Regional Electric Iran Session 3 – Block 3. 4 – Presentation
Barcelona 12 -15 May 2003 Kerman Regional Electric Company & Fault Location Estimation Off-Line Methods l Trial and error method with energization the line section by section On-Line Methods l High frequency transient signals l l Wavelets Pattern recognition l Neural network Mansooreh Zangiabadi Iran Session 3 – Block 3. 4 – Presentation
Barcelona 12 -15 May 2003 Kerman Regional Electric Company Case Study Input data for neural network l l Voltage Current Simulation software l l configuration of feeder and simulator output EMTDC MATLAB Line model l Bergeron model Mansooreh Zangiabadi Iran Session 3 – Block 3. 4 – Presentation &
Barcelona 12 -15 May 2003 Kerman Regional Electric Company & Simulator Output Single-phase to ground fault in the middle of feeder Mansooreh Zangiabadi Iran Three-phase to ground fault in the middle of feeder Session 3 – Block 3. 4 – Presentation
Barcelona 12 -15 May 2003 Kerman Regional Electric Company & The Proposed Neural Network Structure Three-layer feed forward neural network Error back propagation training method Input data – voltage and current – are normalized Output layer Distance l Flag which refers to lateral number l Mansooreh Zangiabadi Iran Session 3 – Block 3. 4 – Presentation
Barcelona 12 -15 May 2003 Kerman Regional Electric Company & The results of L-G fault Training data is prepared in pitches of 50 meters The resistance of fault is changed in steps of 5 from 0 to 25 ohms Mansooreh Zangiabadi Iran Fault Location (m) Fault Resistance (Ω) Error of ANN predicted location (m) Lateral Number 400 1000 5 10 5. 227 7. 439 0 0 1300 15 9. 247 0 1500 5 2. 09 0 1800 10 0. 9 0 2100 15 3. 32 0 2400 10 3. 53 0 2800 5 1. 607 0 Session 3 – Block 3. 4 – Presentation
Barcelona 12 -15 May 2003 Kerman Regional Electric Company Selecting the best structure Number of epochs is considered 1000 epochs Mean Square Error criterion evaluates the structure Error Percentage of Neural Network for L-G fault Mansooreh Zangiabadi Iran Session 3 – Block 3. 4 – Presentation &
Barcelona 12 -15 May 2003 Kerman Regional Electric Company & Structure of Neural Networks as Fault Locator Distance Mansooreh Zangiabadi Iran Session 3 – Block 3. 4 – Presentation Distance
Barcelona 12 -15 May 2003 Kerman Regional Electric Company Combination with GIS Software Mansooreh Zangiabadi Iran Session 3 – Block 3. 4 – Presentation &
Barcelona 12 -15 May 2003 Kerman Regional Electric Company & Thanks for your attention I would also like to thank University of Tehran (UT) and Kerman Regional Electric Company (KREC) for their supports Mansooreh Zangiabadi Iran Session 3 – Block 3. 4 – Presentation
- Slides: 10