The invention discloses a radial-flow type mini-hydro short-period power weighting prediction method. The prediction method comprises steps of: collecting regional radial-flow type mini-hydro data information and non radial-flow hydro data information; calculating the water flow of the ith radial-flow type mini-hydro power station at the jth time, the first short-period power prediction value Pr1ij of the ith radial-flow type mini-hydro power station at the jth time and the first short-period power prediction value Pr1j of the regional radial-flow type mini-hydro power station at the jth time; according to the meteorological data and the output data of the ith radial-flow type mini-hydro power station, carrying out BP neural network prediction to obtain the second short-period power prediction value Pr2j regional radial-flow type mini-hydro power station at the jth time; and determining the weight coefficients of the Pr1j and the Pr2j, and calculating the short-period power prediction Prj OF THE regional radial-flow type mini-hydro power station at the jth time in a prediction day. Thus, technical problems of failure to actually reflect influence on the mini-hydro power stations by multiple factors such as the meteorological conditions, and bigger prediction errors in the current radial-flow type mini-hydro short-period power prediction in the prior art are solved.