The user side has a large number of load resources. The load has uneven capacity, scattered distribution, strong response potential and the ability to participate in grid regulation. Based on the difference in characteristics of power and current during load operation, a fingerprint database of load characteristics is established, and a non-invasive low-voltage load composition identification method for residential appliances based on a multivariate Gaussian model is proposed to achieve online decomposition of residential energy use. Based on the similar characteristics of similar electrical appliances, after obtaining the load actions and interruptible types of the bottom residents, an online aggregation monitoring method for the load demand response capability of the platform area from the bottom to the top is proposed. The REDD data set and the topology of a certain station area are used to test. The results show that the method has a better recognition of the residential load, and can better monitor the capacity of load resources participating in demand response. The method explores a way for participating in the intelligent utilization of system peak shaving and frequency modulation.