Data protection method used in photovoltaic power prediction model training

The invention discloses a data protection method for photovoltaic power prediction model training, which relates to the technical field of data secrecy, and mainly comprises the following steps: receiving global model parameters of each participant in the current round as initialization parameters, training a local model through local data, and obtaining a local data set; local model parameters are uploaded after weighting, encoding and encryption processing of an encryption public key; summing according to the processed local parameter models uploaded by the participants and issuing the sum to all participants in the current round; receiving decrypted content obtained by decrypting the summed local parameter model by each participant through a decryption private key of the participant; summing and decoding are carried out according to the received decrypted contents of all the participants, and the decoded decrypted contents serve as global model parameters of the next round. According to the method, the abu