With the development of power grid a, the traditional operation method of controlling generation to deal with load fluctuation is being challenged. The development of information technology improves the ability of residential load to respond to grid, automatic demand response becomes possible. First, the structure and operational goals of home energy management system (HEMS) are introduced. Then home electrical appliances are classified and modeled to get their operation constraints. Considering that air conditioner is a large controllable load in household, the parameters of the commonly used equivalent thermal parameter (ETP) model are difficult to obtain, which impedes its application. A method of regression is utilized to learn the operating parameters of AC through analyzing historic data. HEMS can optimize operations of electric appliances under real-time electricity price by solving the mixed integer linear programming. Finally, through the example simulation, the effectiveness of AC parameters learning is verified, and scheduling results of the electric equipment are analyzed.