The invention discloses a
power grid fault diagnosis method and belongs to the technical field of
power grid fault diagnosis. An outage region is obtained through protection and
circuit breaker information, lines in the outage region are taken as suspected fault lines, and comprehensive current of the lines is constructed to serve as sample sequences; a DTW
algorithm is used, the DTW distance between the
sample sequence of each suspected fault line and a reference sequence of a
reference line outside the outage area is obtained, and transverse difference degrees are constructed; meanwhile, theDTW
algorithm is used for solving the DTW distance between the sample sequences of the suspected fault lines before and after faults, and longitudinal difference degrees are constructed. The transverse and longitudinal difference degrees of the N suspected fault lines form a 2*N difference
degree matrix; clustering is carried out on the matrix by utilizing an artificial fish swarm clustering
algorithm, the matrix is divided into a
fault class and a normal class, the class with a larger clustering center value is diagnosed as the
fault class, and the lines in the
fault class are judged to be the fault lines. By means of the method, the fault lines can be accurately diagnosed, the influence of fault positions, transition resistance, fault types and the like is avoided, and the performance is good.