Abstract: Load monitoring is an important part of intelligent electricity consumption. A non-intrusive load disaggregation method considering time-phased state behavior is proposed to solve the problem that existing low frequency non-intrusive load disaggregation methods require more priori information and have lower accuracy for load with similar or lower power. Firstly,power data of the load device is clustered to construct a load state template. An intelligent optimization method for the typical behavior time period that does not require a specified number of time periods is proposed. Load state behavior law is extracted by time-phase to construct a load behavior template. Then, on the basis of the traditional power characteristics, considering the two dimensions of probability and time, the time-phased state probability factor（TSPF） is introduced into the objective function as a new load characteristic, and the load disaggregation is realized by multi-feature genetic optimization iteration. Finally, the validity and accuracy of the method are verified on the public data set.