Harmonic Extraction of the Joint Synchronous Extrusion Wavelet Transform and Triangular Basis Function Neural Networks

The parameter identification of time-varying harmonic signals has an important application value. For the frequency of harmonic signals, the synchronous squeezed wavelet transform (SWT) can be used to identify them more accurately, but due to the effect of modal aliasing in the decomposition process, the SWT has a large error in identifying the amplitude and phase values of time-varying harmonics. In order to reduce the recognition error of SWT in harmonic amplitude and phase values and further improve the recognition accuracy, this paper proposes a new method of parameter recognition by combining SWT and triangular basis function neural network. Firstly, the time-varying harmonics are decomposed by SWT, the number of harmonic components is calculated according to the decomposition results, and the estimated value of the frequency of each harmonic component is preliminarily recognized based on the coefficients of SWT decomposition. Then, on the basis of the preliminary recognized values of frequencies, the triangular basis function neural network framework is constructed, and the frequencies, amplitudes and phases of the harmonics are used as the weights of the network, and the learning is carried out in the network by an iterative method. Finally, the network is trained by using the most rapid descent method, and according to the weights of the network after convergence, the frequency, amplitude and phase of the harmonics are converted respectively to realize the accurate identification of the harmonic parameters. The effect of the proposed method is verified through experiments, and the experimental results show that the method in this paper can recognize the frequency, amplitude and phase of harmonics with high accuracy, and when the harmonic signals are contaminated by strong noise, the method in this paper can also accurately recognize the harmonic parameters, and the recognition accuracy is much better than the classical SWT method.