Power load short-term prediction method based on data reconstitution

Disclosed is a power load short-term prediction method based on data reconstitution. The power load short-term prediction method based on the data reconstitution includes: firstly, performing data preprocessing on an initial data set so as to obtain a training sample set; performing the data reconstitution on the training sample set in one day cycle so as to obtain a reconstitution subsequence set; respectively building a time sequence model for each reconstitution subsequence; respectively solving a time series prediction value of a future step under each reconstitution subsequence; performing time series rearrangement on a prediction result so as to obtain a prediction value sequence in a future period. The power load short-term prediction method based on the data reconstitution is strong in operability, and achieves accurate prediction for platform region power loads.