Faulty feeder detection based on mixed atom dictionary and energy spectrum energy for distribution network

The authors proposed a faulty feeder detection method based on the combination of feature extraction and information entropy in the distribution network. Firstly, analysed the characteristics of a transient zero-sequence current (TZSC) waveform, and constructed a mixed atom dictionary by a cosine packet and a wavelet, and then, the authors adopted the matching pursuit algorithm to extract characteristic atoms of each feeder. For energy aspects, used energy spectrum entropy (ESE) to measure; for the waveform similarity, calculated correlation coefficient between characteristic atoms and the original TZSC, then, the authors obtained the ESE of modified atoms and integrated atoms, respectively. In a faulty feeder detection criterion, calculated the difference value between the maximum and the minimum of integrated atoms ESE, and compared the value and the threshold. If the value is smaller than the threshold, the bus is judged as a faulty feeder, otherwise, the maximum corresponding to the feeder is judged as the faulty feeder. Simulation and field tests show that the method can accurately judge the faulty feeder, and its accuracy is immune to transition resistance, iteration numbers, data window length, different compensation degrees, capacitor switching, and unbalanced load.