A spectrum analyzing method based on the autoregressive (AR) model is proposed to analyze the fluctuation characteristics of photovoltaic (PV) power. In order to obtain data with sufficient resolution

the PV array output is simulated based on meteorological data from the National Renewable Energy Laboratory (NREL) of America

which is used to analyze the fluctuation characteristics of PV power stations on different time scales and spatial scales

and determining the time scale on which the photovoltaic power fluctuation component is concentrated. Research shows that PV power fluctuations on different time scales are significantly affected by weather. With the expansion of PV power station scale

the whole output of different PV power stations tends to be smooth. In addition

PV power output fluctuation is mainly concentrated on the 5 min time scale. On this basis

the performance of fitting the probability distribution of PV power fluctuation on the critical time scale with t-distribution is verified. The power fluctuation component on the critical time scale is separated by a rolling average method. Finally

the automatic generation control (AGC) reserve demand caused by PV power grid-connection is quantitatively analyzed. This work is supported by State Grid Corporation of China (No. DZ71-13-038) and National Key Technologies R&D Program (No. 2013BAA02B01).