The stability of the power output of the photovoltaic power plants is affected by the intermittency and uncertainty of direct normal irradiance （DNI）.In order to solve this problem，a prediction model of DNI based on the clustering and rime optimization algorithm （RIME） optimizing the light gradient boosting machine （LightGBM ） is proposed.Firstly，the strongly correlated meteorological parameters of the DNI are determined by the Pearson correlation coefficient，and the historical meteorological data is classified by the mini batch K-means （MBK） clustering algorithm.Then，RIME is used to optimize the hyperparameters of the LightGBM and to establish the prediction model of DNI for different categories of historical meteorological data.The Euclidean distances between the hourly data of the forecast day and the strongly correlated meteorological parameters of each cluster center are used to select the corresponding prediction model for the prediction of the DNI.Finally，by using the historical meteorological data from 2000 to 2019 of a concentrating solar power （CSP） plant in California，USA，the proposed model is validated.The experimental results show that the proposed prediction model can accurately predict the value and variation trend of DNI.