Intelligent measurement technology is an important part of smart grid. In order to realize non-intrusive load monitoring and further improve the accuracy of load identification, a non-intrusive dynamic load monitoring model based on Bayesian network is proposed in this paper, which combines the characteristics of residential electricity consumption behavior and external environment. Gas characteristics and external data are characteristic quantities. Considering the time characteristics of residents and the correlation characteristics of external data, Bayesian network model is used to model and analyze residential electricity consumption behavior, and the feature database is updated dynamically over time, so as to realize the monitoring function of household load. This paper uses AMPds2 public data set data to verify the algorithm, which proves the accuracy and validity of the algorithm. Mutual information analysis of external data and electricity consumption behavior shows that the time-period characteristics have the strongest correlation with electricity consumption behavior.