The high randomness and volatility of photovoltaic (PV) power makes it difficult for single prediction models to accurately analyze the fluctuation patterns in historical data,resulting in low prediction accuracy.To address this issue,a combined model for short-term PV power prediction was proposed,which incorporated Convolutional Neural Network,Gated Recurrent Unit (CNN-GRU) and an Improved Sparrow Search Algorithm (ISSA) for optimizing the eXtreme Gradient Boosting (XGBoost) model.First,the historical data were normalized after outlier removal,and feature selection was carried out via Principal Component Analysis (PCA) so as to better identify the key factors affecting photovoltaic power.Then,the CNN and GRU networks were used to extract the spatial and temporal features of the data,respectively.To address the difficulty in manually configuring parameters and high randomness of the XGBoost model,ISSA was used to optimize the hyperparameters of the model.Finally,the reciprocal error method was used to reduce the error of the results predicted by the two methods (CNN-GRU and ISSA-XGBoost) while the weights were updated to obtain new predicted values to complete the prediction of photovoltaic power.The experimental results show that the proposed CNN-GRU-ISSA-XGBoost model has strong adaptability and high accuracy.