Short-term fluctuations of solar power output via cloud shadows are one of the factors causing difficulty in predicting the output of solar power generation. In this paper, the short-term fluctuations of solar power output are analyzed using a discrete Fourier transform, short-time Fourier transform, and continuous wavelet transform. Based on the frequency analysis results, the necessity for short-term predictions are presented. It was found that continuous wave conversion is most suitable for the analysis of non-stationary waveforms such as the output of solar power generation. From the scalogram obtained by the continuous wavelet transform of the output power waveform of solar power generation, the frequency band corresponding to the short-term fluctuation was determined. The short-term fluctuation time was calculated by applying envelope processing to the absolute values of the spectral intensity in the frequency band, and it accounted for ~29% of the annual solar power generation time. Therefore, it was clear that approximately 29% of the annual solar power generation time requires a power generation prediction interval as fast as the floating speed of cloud shadows.
