Fig 1.
Discrete wavelet transform decomposition process.
Fig 2.
Basic artificial neural network architecture.
Fig 3.
Extracted coefficients using Db4 mother wavelet.
(A) Original voltage waveform. (B) Approximation coefficients. (C) Detail coefficients.
Fig 4.
(A) Enlarged approximation coefficients. (B) Enlarged detail coefficients.
Fig 5.
Flowchart of ANN optimization algorithm.
Fig 6.
Simplified distribution network.
Fig 7.
Fault impedance estimation result between standard ANN and PSO-optimized ANN.
Fig 8.
Fault distance estimation result between standard ANN and ANN-PSO.
Table 1.
ObjFunc based on different type of learning algorithm.
Table 2.
Robustness of trainbr and trainlm algorithms.
Fig 9.
Consistency and competency between trinlm and trainbr.
Table 3.
Fault impedance estimation using PSO-optimized ANN (18-bus).
Table 4.
Fault distance estimation using PSO-optimized ANN (18-bus).
Fig 10.
Distribution network consists of 33 buses.
Table 5.
Fault distance estimation (AGF).
Table 6.
Corrected fault distance estimation (AGF).
Table 7.
Fault distance estimation (BGF).
Table 8.
Fault distance estimation (CGF).
Table 9.
Fault distance estimation (ABCGF).
Table 10.
Fault distance estimation (ABCF).
Table 11.
Fault impedance estimation (AGF).
Table 12.
Fault impedance estimation (BGF).
Table 13.
Fault impedance estimation (CGF).
Table 14.
Fault impedance estimation (ABCGF).
Table 15.
Fault impedance estimation (ABCF).
Table 16.
Comparing the proposed method to existed methods.