The invention provides a feeder baseline load prediction method. The method comprises the steps that historical meteorological data, the historical load data of all the users on a feeder and the meteorological data of a prediction period are imported; the historical load data of the previous year are corrected; the historical days of similar meteorology are selected; baseline load prediction of all the users on the feeder is performed by using an RBF radial basis artificial neural network method; the baseline load prediction values of all the user of the feeder are adjusted by using an adjustment factor; and summation of the baseline load prediction results of all the users of the feeder is performed. The influence of the meteorological factors on the load is considered in selection of the historical days of similar load, and the holidays and festivals, the demand response event days, the reservation cutting days and the days of which the load is lower than the threshold are eliminated so as to ensure the rationality of selection of the similar historical days. According to the feeder baseline load calculation method, the radial basis artificial neural network is selected, and the radial basis artificial neural network has the advantages of fast convergence rate and can easily search the global minimum in comparison with a BP neural network commonly used in electric power system load prediction.