Abstract:The grey relational analysis is conducted to determine the meteorological environment factors with the highest impact on photovoltaic power generation. Then the key factors to construct a model of forecasting short-term photovoltaic power generation with the multi-variable time series CAR (Controlled Auto-regression) model based on the historical daily data. Finally, the output of photovoltaic power station is predicted. The prediction experiment is conducted based on the operational data of 18 kW grid-connected PV plant in the Power Electronics Research Center of Huazhong University of Science and Technology. An appropriate model is obtained by adjusting model parameters. The experiment results verify the validity of this method, and indicate that the precision of the prediction is relatively high in a good weather condition.