Intelligent measurement technology is an important part of smart grid. In order to enhance the applicability of the non-invasive family load identification algorithm, this paper proposes a low frequency monitoring and combined with residential electricity load behavior associated with external characteristic of power load characteristic, to build a family load monitoring model based on random forest. In this model, firstly, the commonly used electrical characteristics as well as the introduction of external data such as the time characteristics of resident load characteristics, through the analysis of the mutual information method selection and multi-dimensional characteristics of electricity behavior correlation is high, and the random forest algorithm is adopted to residential electricity behavior modeling and load monitoring, So as to realize the effective monitoring of different types of load in different families. Finally, the algorithm was run on the AMPds open data set and compared with the bayesian classification algorithm, and the results verified the effectiveness of the proposed algorithm.