The invention discloses a transformer substation acoustic signal feature extraction method based on MFCC, and belongs to the field of monitoring. The method comprises the following steps: performing acoustic signal preprocessing, fast Fourier transform, Mel filter bank and logarithm DCT operation on a transformer acoustic signal in sequence, and extracting the characteristic quantity of the transformer acoustic signal; wherein the preprocessing comprises two steps of framing and windowing; converting the time domain signal into a frequency domain through fast Fourier transform, and calculatingan amplitude spectrum and spectral line energy of each frame; enabling the Mel filter bank to calculate energy for each frame of spectral line energy spectrum through the filter bank; carrying out logarithm DCT operation to obtain logarithms of the energy values, calculating an energy matrix and identifying the fault type in a fuzzy clustering or neural network mode. Distinguishing is carried outthrough timbre, extracted feature vectors have obvious difference, the operation speed is high, and the operation cost of a computer can be reduced; and the feature vectors are classified and markedby using a support vector machine algorithm, so that discrimination speed is high, accuracy is high and instruction is simple. The method can be widely applied to the field of operation monitoring andfault judgment of unattended substations.