Grid Harmonic Detection Based On WNN And EEMD Methodological Study

A large amount of non-linear load is connected to the grid,which brings harmonic pollution to the grid.People's pursuit of high quality of life has made the control requirements of power grid harmonics higher.The key to harmonic control lies in the real-time detection and high-precision extraction of harmonics.This is because the information obtained by real-time detection and high-precision extraction provides input information and feedback information for high-quality harmonic control.Since the harmonics of the power grid are nonlinear,the conventional harmonic detection method is difficult to meet the requirements of real-time harmonic detection and high-precision extraction.Therefore,it is of great significance to study harmonic detection methods with good real-time and high accuracy.The basic principles of wavelet analysis and the basic properties of common wavelet functions are analyzed.The multi-resolution analysis and Mallat algorithm in wavelet analysis are introduced,and its application in detecting harmonics is listed.Then the classical BP neural network is selected as the research object,and its network structure and training algorithm are analyzed.The application of neural network in the harmonic detection of power grid is given,which lays a theoretical foundation for the later research.A wavelet neural network adaptive optimization method is proposed.This method is aimed at the problem that the network initial value is improperly set and the network convergence is slow or even not convergent.A parameter adaptive optimization adjustment method is given.In the aspect of network training,the training algorithm with additional momentum items smoothes the weight learning path,effectively avoids network training falling into local minimum,significantly improves network performance,and has fast convergence speed,which can effectively improve the real-time performance of harmonic detection.Compared with other detection methods,it is proved that the proposed method has fast convergence and good real-time performance.An improved EEMD algorithm is proposed.This method is aimed at the EEMD decomposition process,due to the addition of a single white noise signal,it will inevitably affect the modal function,and propose to add white noise signals with opposite positive and negative amplitudes to eliminate the drawbacks.Schmidt orthogonalization theory orthogonalizes each modal function to avoid modal aliasing and improve harmonic decomposition accuracy.The simulation proves that the improved method can effectively suppress the modal aliasing phenomenon and reduce the orthogonalization index between the modal functions,thus improving the decomposition accuracy,which is extremely beneficial for identifying and extracting harmonic components.