Abstract: In response to the current situation where the accuracy of solar power prediction in existing photovoltaic stations cannot meet the needs of power grid dispatch, this article uses historical meteorological factors and solar power data to first calculate their correlation coefficients and extract meteorological factors with significant impact. Secondly, perform grid optimization on the meteorological factors output by the grid based numerical forecasting model. Finally, compare the accuracy and benefits of optical power prediction before and after optimization processing. Research has shown that there is a significant correlation between relative humidity, temperature, and light power. The average relative error between the optimized numerical forecast and the measured temperature is 1.74 ℃ and 0.8 ℃, respectively. The optimized model can significantly improve the accuracy of meteorological factors. The calculated values of optical power before and after optimization are close to the measured power. At the power peak position, the calculation accuracy has increased from 76% before optimization to 95% after optimization.