The invention discloses a super-long-term prediction method for a super-rich water year of reservoir incoming water. The method comprises the following steps: respectively taking the relative number of sunspots as an X axis and a moon declination angle as a Y axis; using the date of recent days as an X axis, and using the moon declination angle as a Y axis; taking the date of recent days as an X axis, and taking the relative number of sunspots as a Y axis; drawing first to third composite factor point aggregation graphs; based on a clustering effect, respectively marking out a super-rich wateryear point aggregation area on the three composite factor point aggregation graphs; obtaining a sunspot relative number, a moon declination angle and a recent date of the year to be predicted; finding out an extra-rich water year point aggregation area to which the year to be predicted belongs on the three composite factor point aggregation graphs respectively, and finally processing historical year incoming water data with high occurrence frequency in the extra-rich water year point aggregation area to obtain incoming water data of the year to be predicted. Ultra-long-term prediction is carried out on the reservoir inflow ultra-rich water year based on the multiple astronomical factors, and compared with an existing prediction method based on a single astronomical factor, the predictionprecision is improved.