Power network digital power flow prediction method

The invention relates to a power network digital power flow prediction method, which solves the problem of aliasing interference of multiple intrinsic mode components of a flow sequence, reduces the operation complexity of an LSTM neural network, shortens the operation time and reduces the prediction error by performing VMD decomposition preprocessing on the flow sequence compared with a traditional RNN, LSTM and other neural network flow prediction methods. Meanwhile, regression analysis is carried out through an SVM regression method, prediction deviation caused by sudden inrush flow of data flow is effectively reduced, and the precision of the model is further improved; and fitting each component prediction value of the flow sequence passing through the LSTM neural network model with a residual component regression analysis value to obtain a VMD-LSTM-SVM model electric power digital power flow accurate prediction value. Therefore, the power communication network can make a scheduling response to possible c