Non-intrusive load monitoring(NILM) can be applied in smart grids, it can playing an important role in energy saving and emission reduction, supply and demand side management, and optimization of grid structure. By analyzing the characteristics of the household electricity data, the amplitude and slope change of the active power are selected as the input of the neural network, and the use information of the household appliance is output. Train and validate with public data sets. The results show that the method can effectively realize the resident load identification.