System frequency response model modeling method based on AM-LSTM neural network

The invention discloses a system frequency response model modeling method based on an AM-LSTM neural network, and belongs to the field of power system frequency response model modeling. Aiming at themodeling problem of a wind-fire coupling system frequency response model, the invention provides a system frequency response model modeling method based on the combination of an attention mechanism and a long-short-term memory neural network, and the method comprises the steps: extracting the time sequence features inputted by the coupling system frequency response model through the attention mechanism; the wind-fire coupling system can be effectively utilized to input data; the long-short-term memory neural network is utilized to solve the problem of gradient disappearance of the recurrent neural network in the updating process, and in addition, the long-short-term memory neural network solves the problems that an existing coupling system frequency response model is difficult to describenonlinearity, uncertainty