The present invention relates to a power distribution network dynamic reconstruction method based on data drive. The method comprises the steps of: extracting symbol features of a daily load curve, performing rough matching of a plurality of historical load curves as a first candidate load sequence; taking a time frame division method of the historical load curves being the most similar to the current daily load curve as a time frame division method of the current daily load curve; performing rough matching of a plurality of historical load sequences as a candidate load sequence as a second candidate load sequence; performing dimension reduction of original features, determining weights of the features after dimension reduction, performing fine matching of the second candidate load sequence, wherein each time frame matches one candidate load sequence being the most similar to the current sequence; and taking a static reconstruction method as a static reconstruction method of each timeframe of the current daily load curve. The defects are solved that a traditional method needs mutual setting of parameters and is easy to converge the locally optimal solution, and the power distribution network dynamic reconstruction method based on data drive greatly reduce the computing time of reconstruction and is specially suitable for the real-time optimization dispatching of the power distribution network.