Improved random algorithm for new energy power generation prediction error

The invention discloses an improved random algorithm for a new energy power generation prediction error. The improved random algorithm comprises the following steps of 1, inputting grid-connected power data, meteorological substation data and numerical weather forecast data; 2, processing the input data, inputting gridding numerical weather forecast data, outputting cluster power, constructing a meteorological group and a factor by using a data mining means, constructing a new energy prediction error random model, and achieving the overall prediction of the regional new energy power generation cluster power; 3, based on the real-time operation data, through weather situation research and judgment and transverse and longitudinal prediction deviation morphological feature mining and correction, updating an original prediction result in a rolling manner, and returning a new prediction result in real time; the method has the advantages that modeling efficiency is high, compared with an original off-line modeling-