Method for predicting load of electric vehicle under condition of data shortage

The invention discloses an electric vehicle load prediction method under data shortage, and the method comprises the steps: enabling a conventional GAN to learn a potential relation between observation values containing irregular time lag through employing a GRUI cell structure for the irregular time lag change between front and back effective observation values in EV load data caused by a vacancy value; data restoration is carried out through an interpolation method adapting to EV load data, and a data set is obtained; the performance of the LSTM network is improved by adopting a Mogrifier gating mechanism, and an EV short-term load prediction result is obtained on the processed data set; the technical problem of low EV load prediction precision caused by data missing and data abnormity in the prior art is solved. 本发明公开了一种数据缺乏下电动汽车负荷预测方法，所述方法为：针对由于空缺值所导致EV负荷数据中前后有效观测值之间的不规则的时滞变化，采用GRUI细胞结构让传统GAN学习到包含不规则时滞的观测值之间的潜在联系，从而适应EV负荷数据的插补方法来进行数据修复，得到数据集；采用Mogrifier门控机制提升LSTM网络性能在经处理过的数据集上获取EV短期负荷预测结果；解决了现有技术数据缺失和数据异常