The invention discloses a GCM correction method of considering day-by-day data fluctuation features. Months are used as calculation units to extract day-by-day actually-measured data of a historical period of a certain meteorological factor, day-by-day data of GCM of the same historical period and to-be-corrected day-by-day data of the GCM of a future period; multi-year monthly average values of the historical actually-measured data and the historical GCM data are calculated; average-value correction factors are constructed; thus day-by-day average-value correction is carried out on the futuremeteorological data; variance correction factors are constructed; and variance correction is carried out on the future meteorological data after average-value correction, and variance after correction is used as constraint conditions to obtain a new future GCM meteorological-sequence. According to the method, consistency of distribution features of the future GCM data and distribution features ofthe actually-measured data is considered, deviation correction is carried out on the future GCM data, and thus results obtained by simulation of a model adopting the GCM data are enabled to be more accurate.