The invention belongs to the field of electric power system power distribution network fault diagnosis, and particularly relates to a power distribution network fault diagnosis method integrating an electromagnetism-like mechanism in a temporal ambiguity Petri network, so that more accurate element fault probability is obtained by modifying parameters such as initial confidence coefficient and weight. The method includes the steps: 1 acquiring alarm information of all-level protection and circuit breakers received by a SCADA (supervisory control and data acquisition) system after faults happen; 2 searching a power supply interrupted region by a power topology analysis method to obtain a suspicious faulted element set in a faulted region. Fault diagnosis range is narrowed, and a set of all suspicious elements and corresponding alarm is built.