Transformer substation noise separation method

The invention provides a transformer substation noise separation method and relates to the technical field of audio processing. The technical purpose that three types of noise of a high-voltage transformer substation is independently measured is achieved. The method includes the steps that firstly, multilayer wavelet transformation is conducted on noise of the high-voltage transformer substation, so that six groups of wavelet coefficients are acquired; then, after the low-frequency wavelet coefficients pass a passband comb filter, wavelet inverse transformation is conducted, and therefore body noise is separated; meanwhile, after the low-frequency wavelet coefficients pass a stop band comb filter, wavelet inverse transformation is conducted on the low-frequency wavelet coefficients and other five groups of wavelet coefficients, and mixed signals containing corona noise and fan noise are acquired; then the corona noise is detected through the high-frequency wavelet coefficients, a fan noise power spectrum is estimated in a period without the corona noise, a corona noise power spectrum is estimated in a period with the corona noise, time domain signals are reconstructed through the estimated fan noise power spectrum and the estimated corona noise power spectrum, and therefore the fan noise and the corona noise are separated. The method is suitable for processing noise signals of large high-voltage alternating current and direct current transformer substations.