The invention provides a non-intrusive load decomposition method and system based on a long-short-term memory network. The method comprises the following steps: collecting a plurality of total load data in a period of time, and carrying out differential processing and window sampling processing; Inputting the total load data subjected to differential processing and window sampling processing intoa pre-established composite deep long-short-term memory network model to obtain a power value of each electric appliance type; Wherein the composite deep long short-term memory network model comprisesthe relationship between the total load and the power value of each electric appliance type. According to the method and the system, the problems of high hardware cost and large data volume of high-frequency sampling data under the intrusive load decomposition research are avoided, and the power of a small electric appliance, a continuously changing electric appliance and an electric appliance which is used all the time can be decomposed.