The fluctuation of PV system output caused by meteorological changes would threaten the voltage stability and operation control of the power grid in the area of high PV permeability. Therefore, a meteorological data acquisition terminal with STM32F103 as the core processor is designed in this paper, and the measurement data is uploaded to the data integrated processing unit composed of nginx server and Apache server by GPRS and HTTP transparent transmission module. Among them, Nginx server is responsible for data access and distribution, and Apache server is responsible for data analysis and processing. Finally, a NARX neural network model is constructed, and the short-term prediction of solar radiation is studied by using the measured data. The results show that the proposed scheme can effectively predict the short-term changes of solar radiation in the surrounding areas of photovoltaic power plants, and provide effective support for the prediction of power changes of new energy generation and the optimal operation control of distribution network.