In the wake of global warming and energy shortage,the smart grid has become the goal of the power grid development,for which intelligent household appliance network system is very important. However,it is difficult to combine conventional household appliances with new system. To deal with the problem,this paper focuses on the appliances recognition algorithm applicable to traditional and typical household appliances. By using sequential data of appliances power from intelligent sockets,the paper generalizes and extracts the characteristics of the occupant behavior and the power of typical household appliances. A new recognition algorithm of household appliances is presented and realized on the basis of Bayes classification model in large data. Seven types of household appliances (refrigerator,electric cooker,television,laptop computer,washing machine and water fountain) in 15 households in Beijing were tested. This algorithm is proved to possess higher accuracy,lower cost and applicability.