The invention relates to a transformer collision record data analysis method based on a wavelet energy method. The transformer collision record data analysis method is characterized by comprising thesteps of: setting a sampling frequency of a collision recorder, and recording acceleration data in three-dimensional space directions at different time points; exporting the acceleration data in the three-dimensional space directions, calculating a resultant acceleration sequence and acquiring a time period during which severe collision possibly occurs; performing multi-level wavelet decompositionon a resultant acceleration by using a db4 wavelet, so as to obtain an approximation signal and a detail signal; and carrying out comprehensive judgment according to signal energy proportion of all levels and the maximum acceleration, and determining whether a transformer suffers severe collision. The transformer collision record data analysis method realizes the function of analyzing collision and vibration signals during the transportation process of the large-scale power transformer, eliminates interference of high-frequency signals during transportation, can effectively evaluate influenceof factors such as jolt and collision during the transportation on equipment, improves the effectiveness of data analysis of the impact recorder, and effective assists the transport personnel in improving the precision of collision data analysis.