The invention relates to a transformer excitation surge current identification method based on time-frequency characteristic quantities, and belongs to the technical field of transformer relay protection. The method comprises the steps: if the differential current of a transformer is larger than a setting value, decomposing data inside continuous N sliding time windows by utilizing the discrete wavelet decomposition, respectively calculating the total energy Ej of differential currents inside the sliding time windows and the energy sum Edi of frequency bands of the sliding time windows, obtaining the energy percentage Eij of each sub frequency band inside one time window, and forming a characteristic matrix WTF. A correlation coefficient of integrated time-frequency characteristics of the sum inside two adjacent time windows is calculated. When is smaller than , the is equal to 1, or the is equal to 0. By utilizing the correlation coefficient for representing the time-frequency characteristics of different current signals and the change rules of the different current signals, the integrated correlation coefficient is combined to form an excitation surge current identification criterion. According to the transformer excitation surge current identification method based on the time-frequency characteristic quantity, by starting from the analysis of the differential current time domain and the frequency domain, the time-frequency characteristic differences between fault currents and excitation surge currents inside different sliding time windows are effectively found out, and compared with a traditional surge current identification method by utilizing a single characteristic quantity, the transformer excitation surge current identification method based on the time-frequency characteristic quantities has the advantages of integrating multiple pieces of information such as the amplitude, the phase, the singularity and frequency distribution of the differential current and having higher reliability.