Aiming at the matching characteristics of appliances in non-intrusive load monitoring, this paper proposes a matching method based on Hungarian algorithm. This method using Prony sliding window algorithm to detect the event, and therefore extracts the characteristic change of the load information. The proposed method converts the power variation characteristics of appliances into a bipartite graph optimization matching problem and combines the augmented path to find the perfect match when appliances are turned on and off. Furthermore, in order to avoid the power of the load event being turned on and off is not equal, the algorithm is improved by adding virtual nodes. Also, we introduced grey correlation evaluation and multiple matching strategies in this paper. Experimental results show that the proposed method can identify the opening and closing of the load effectively, and lays a foundation for improving the accuracy of load identification.
