The invention discloses a long-range year typical daily load prediction method based on AP clustering, and the method comprises the steps: firstly selecting historical daily load data, and carrying out the per-unit processing; secondly, clustering the historical daily load data based on an AP clustering algorithm, and taking the historical daily load data as load characteristics of a long-range year; and finally, calculating a typical daily load curve of the long-range year based on the predicted value of the load of the long-range year in combination with the load characteristics after clustering. The method provides an effective scientific basis for prediction of the daily load of the long-range year, the daily load of the long-range year is clustered to obtain the typical daily load ofthe long-range year, and the typical daily load is used for calculation, so that the calculation amount in power grid planning and power grid operation is reduced.