an intelligent approach using svm to enhance turn to turn
play

An Intelligent Approach using SVM to Enhance Turn-to-Turn Fault - PowerPoint PPT Presentation

An Intelligent Approach using SVM to Enhance Turn-to-Turn Fault Detection in Power Transformers Dr. Mohamed Elsamahy, P.Eng. Department of Electrical Power & Energy Military Technical College, Cairo, Egypt Mariya Babiy, E.I.T Department of


  1. An Intelligent Approach using SVM to Enhance Turn-to-Turn Fault Detection in Power Transformers Dr. Mohamed Elsamahy, P.Eng. Department of Electrical Power & Energy Military Technical College, Cairo, Egypt Mariya Babiy, E.I.T Department of Transmission Development & operations planning ATCO Electric, Edmonton, AB, Canada

  2. Contents  Motivation  Support Vector Machines (SVM)  SVM proposed scheme for transformer turn-to-turn faults detection  Simulation results  Conclusions 2

  3. Motivation “Relay (87) lack of efficiency during turn -to- turn faults” • According to IEEE failure statistics of power transformer during turn- to-turn faults:  94 power transformers failed “ 1997-2001 ” .  50% of these failures are winding failures • According to IEEE Standards there is no one standard way to protect all power transformers against minor internal faults and at the same time satisfies the basic protection requirements: sensitivity, selectivity, and speed. • According to IEEE Standards much as 10% of the transformer winding might be shorted to cause a detectable change in the terminal current. • Therefore the need for new protective techniques to overcome the previous difficulties has been increased …………… (SVM) 3

  4. Transformer Failure due to Shorted Turns (Case Study-Euro TechCon 2009, UK) • Date: April 2009 • Service life: 30 years • Rating: 750 MVA, 400/275/13 kV • Transformer was tripped on Buchhloz relay • Fault diagnosis: -Severe shorted turns in the middle windings in the middle phase. -Extensive loss of conductors and conductor insulations in the upper part. 4

  5. (Case Study-Euro TechCon 2009, UK) Fig.1 5

  6. Support Vector Machines (SVM) Linearly separable data Non-separable (noisy)data Class A H H ζ i H 1 Class A i H 2 H 2 ζ i H 1 Class B m r Class B m r Class A Feature map Non-Linear data Class B Fig.2 Complex in low dimensions Simpler in high dimensions 6

  7. Support Vector Machines (continue) Class A Non-Linear Feature map data Class B Fig.3 Complex in low dimensions Simpler in high dimensions (n) is the polynomial order ( C ) is the penalty due to error C ( γ ) is the Gaussian width 7

  8. SVM proposed Scheme Y-Y Other parts Other parts of the system of the system V, I Fig.4 Feature Vector ( x ) x (20 samples per cycle) - Sampling frequency of 1.0 kHz. SVM_D y Internal turn-to-turn Fault (+1)/otherwise (-1) - A data window of one-half cycle (form the fault inception time) is used for internal turn-to-turn fault detection. 8

  9. Simulation results (training and testing data for SVM proposed module)

  10. Simulation results (continue) G1, T1 Training and testing data Black ≡ Training data (during turn-to-turn faults) 2 % 1 % Red ≡ Testing data 6% 3 % 12% 5 % 18% 7 % 60 % 55% 25% 10% 30% 0 2 Ω 70 % 65 % 15% 1 Ω 6 Ω 20% 5 Ω 25% 80 % 75 % 10 Ω Number of shorted turns Fault resistance % of transformer windings under 85 % 90 % test Transformer loading G1 size = 4 × 4 × 7 × 2 = 224 % of Transformer MVA T1 size = 4 × 2 × 6 × 2 = 96 rating 9

  11. Simulation results (continue) G3, T3 Training and testing data G2, T2 Training and testing data (during energization inrush) (during external faults) 0 2 Ω 1 Ω 6 Ω at 3 s 60 % 55% 5 Ω 60 % 55% for 50 m.s 10 Ω 70 % 65 % duration Fault resistance 70 % 65 % 80 % 75 % B-g 80 % 75 % G3 size = 4 B-C 90 % 85 % T3 size = 4 85 % 90 % Fault type Transformer loading % of Transformer MVA G2 size = 64 Transformer loading rating % of Transformer MVA T2 size = 32 rating 10

  12. Simulation results (continue) Finally, the training data set (Set_1) size = 292 samples [G 1 (4 × 4 × 7 × 2 = 224) + G 2 (4 × 4 × 2 × 2 = 64) + G 3 (4)] while the testing data set (Set_t) size = 132 samples [T 1 (4 × 2 × 6 × 2 = 96) + T 2 (4 × 2 × 2 × 2 = 32) + T 3 (4)] 11

  13. Simulation results (continue) SVM scheme design 292 cases 132 cases n = 10 n = [2,10] n = 8 step 2 C = 1, 10, 100, C = 1000 500, 1000 C = 500 γ = 0.1 γ = 0.1, 0.2, γ = 0.2 0.3, 1, 3, 5 12

  14. Simulation results (continue) Table.1 Performance efficiency of conventional differential relay and SVM technique during internal turn-to-turn faults for all operating conditions Generator Phase Backup Protection Performance Efficiency Relay (87) 78.47% Proposed SVM scheme (Polynomial Kernel) 97.72% 98.48% Proposed SVM scheme (Gaussian Kernel) Where, No . of correct tripping    % 100 Total No . of cases 13

  15. Conclusions • Support vector machines (SVM) classification technique is reliable with high performance efficiency for transformer turn-to-turn fault detection (average performance efficiency of 96.2% and detection time within one-half cycle form fault inception time (8.33 m.s). • In comparison to the performance of Relay (87) the proposed scheme reflects appreciable enhancement in the transformer protection against internal turn-to-turn faults. 14

  16. THANK YOU For further information, please contact Dr. Mohamed Elsamahy, P.Eng. Department of Electrical Power & Energy Military Technical College, Cairo, Egypt mohamed.elsamahy@usask.ca mohamed.elsamahy@ieee.com

Recommend


More recommend