Power system recovery decision-making method based on LSTM neural network

The invention discloses a power system recovery decision method based on an LSTM neural network, relates to the technical field of power system recovery, and solves the problems that photovoltaic prediction is not accurate enough and decision output is not perfect enough in power system recovery. According to the method, firstly, a photovoltaic output prediction LSTM network is trained through collected data such as time, irradiance, humidity, air pressure, actual photovoltaic output power and the like, power output data in a future time period is accurately predicted based on the photovoltaic output prediction LSTM network, and output of a power system recovery decision is performed based on the predicted data. And the hysteresis of the power system recovery decision is eliminated. In the power system recovery decision output process, the maximum recovery load capacity serves as a target function, constraint conditions comprise unit state constraint, load state constraint, power system power balance constrai