Abstract: Perceiving state and identity of real-time electrical load is the premise of refined energy management and demand response. To meet the automatic and precise control requirements of smart electric appliance network, the on-line identification technology of appliance-level load state and type is proposed. On the basis of real-time measurement and communication of multi-appliances, the state of load is extracted by improved CUSUM segmentation and hidden Markov model, and the single classification method SVDD is used to identify the electrical appliance type extensively. The accuracy, timeliness and extensive ability of the proposed methods are verified by constructing smart electric appliance network in residential and office buildings.