Load switching events are important basis for load classification and identification. In order to accurately identify the process of load switching, a non-intrusive load identification method based on KM matching algorithm is proposed in this paper. An approximation method fitting the power curve is used to detect load events, and the probability distribution model of load event signatures is established by using the switching steady-state signatures. Considering that the number of load turning on and off events is not equal, the load events are matched with loads in database, and weighted optimization KM algorithm is used to find the optimal solution, so as to realize the correct load matching. In the test of real scenario and REDD dataset, the results show that the proposed method can effectively match and identify load events, which lays foundation for the subdivision of subsequent energy consumption.