Abstract: A photovoltaic power prediction model based on quadratic decomposition and improved particle swarm optimization algorithm is proposed considering the impact of different solar radiation on photovoltaic power. Through Spearman and Kendall’s correlation analysis of various meteorological factors affecting photovoltaic power, it was found that the correlation coefficients between total tilt radiation, total horizontal radiation, diffuse tilt radiation, diffuse horizontal radiation, and photovoltaic power are relatively large. Then we use CLARANS to divide the sample data into strong radiation, medium radiation and weak radiation according to the solar radiant intensity. For the three types of data, we use CEEMDAN to decompose the key meteorological factors and power twice, fully mining time series information and reducing data instability. The GWCPSO is proposed to optimize the hyperparameter of the convolutional neural network and the bidirectional long short-term memory network, improve the efficiency of parameter adjustment, and finally build a prediction model for photovoltaic power prediction. Analyzing the prediction errors of different decomposition methods and network models under three types of solar radiation, the results show that the proposed prediction model can effectively improve the prediction accuracy of photovoltaic power under different solar radiation conditions.