Power electronic noise energy modeling method based on multi-scale wavelet transform network prediction

The invention relates to a power electronic noise energy modeling method based on multi-scale wavelet transform network prediction, which comprises the following steps: aiming at an operation scene of a power electronic system, after power electronic noise is collected, analyzing time-frequency characteristics and distribution characteristics of the noise, and adopting multi-scale wavelet transform to obtain an excitation function; according to the method, a deep neural network is utilized to perform noise parameter adjustment and weight updating, an optimal weight is obtained through parameter prediction and model optimization, and modeling is performed for different noise types, so that the reliability and stability of noise modeling are greatly improved, and the data transmission efficiency of an electric power communication system is improved. Besides, by adjusting scale factors in a multi-scale wavelet transform network prediction algorithm, frequency values of multi-noise features are obtained, and freq