In order to contain fault information as much as possible and improve the accuracy and robustness of fault line selection based on the transient feature, we propose a novel fault line selection method based on sparse decomposition and comprehensive measure for distribution network. Firstly, when fault occurs, matching pursuits method is introduced to decompose the transient zero sequence current of the first 2 cycles, and the iterations are set to 4. Secondly, the best atoms of first 1/4 cycle are selected as the study object, on the one hand, the atomic singular entropy is computed, on the other hand, the feature atoms is obtained after reconstructing best atoms, and the actual energy of feature atoms are calculated. Finally, the fault confidence function is constructed, whose independent variable is the reciprocal of atomic singular entropy after normalizing. Thereby, the fault comprehensive measurement function is built, by which the fault information can be quantified in different perspectives. And the line with the maximum of fault comprehensive measure is chosen as the fault line. It can be known from the simulation results that, when the fault occurs, the obtained comprehensive measure of fault line is larger than that of the healthy lines, the selection results based on this feature are not affected by fault line types, fault angles, grounding resistors, and fault locations. © 2017, High Voltage Engineering Editorial Department of CEPRI. All right reserved.
