The invention provides a lateral error translation modification method of a wind electricity power forecast system. The method includes the steps that wind electricity power historical forecast and actual measurement data sequences are collected and are classified according to four seasons, and data of each season every year are divided into three kinds of power trend sections, that is a continuous power increasing section, a continuous power lowering section and an unchanged power section; duration time sequences of each power trend section are obtained according to power trend duration models, probability statistic analysis is carried out to obtain a macroscopic centralized distribution value, a wind electricity power forecast lateral error model is input to obtain a lateral error of each season every year, and a weighted average value of the lateral errors of each identical season of N years serves as a lateral error modification value of each season; translation modification is carried out on forecast power sequences supplied to power grid dispatching by the wind electricity power forecast system by means of the weighed average value, and the translation modification effect is evaluated on the whole. The method has common adaptability for power short-term or exceedingly-short-term forecast systems of wind electricity, photoelectricity and the like and can further improve precision of various power forecast systems.