Distributed power supply output prediction method and device based on combined neural network

The invention relates to the technical field of distributed power supply output prediction, in particular to a distributed power supply output prediction method and device based on a combined neural network, and the method comprises the steps: constructing CNN units of a corresponding number according to the number of output influence factors, and enabling each CNN unit to be in parallel with an attention branch, thereby completing the construction of CNN-attention units; all CNN-attention units are paralleled, and each CNN-attention unit is in butt joint with an LSTM (Long Short Term Memory), so that the construction of a distributed power supply output prediction model is completed; acquiring historical output data of the distributed power supply; training the distributed power supply output prediction model according to the historical output data; and obtaining latest output data of the distributed power supply, and performing output prediction of the distributed power supply through the trained distribute