The invention provides a state matrix symmetry-based power distribution network abnormal data identification method. The method includes the following steps of: determining the state quantity of a reactive power optimization input state matrix; collecting sample data and obtaining the probability distribution of the sample data; and constructing a probability distribution-based expected state matrix; and using the symmetry of Hermitian matrix to test the abnormal data of the reactive power optimization input state matrix. With the method adopted, the abnormal data can be quickly and accurately identified, and global optimum can be achieved.