Harmonic Analysis Approach Based On Wavelet Transform And Neural Network In Power System

Harmonic is an important aspect of power quality, and harmonic analysis and measurement occupies an essential position in power system. The existence of harmonic can not only endanger the safety of power facilities but also affect the stability of power system. In recent years, the application of power electronic devices improves the electrification and automation level, but it also makes the voltage and current distortion more serious. This also causes the increase of harmonic content and complexity. Therefore, it is obligatory to solve this issue. Harmonic analysis is the core and foundation of harmonic research. This paper presents a harmonic analysis method based on wavelet transform and artificial neural network which makes a preliminary theoretical exploration for the high-precision establishment algorithm in real-time.Firstly, the paper briefly describes the basic concepts and harm of harmonic. The current main harmonic analysis methods are compared. Then it separately proposes the basic theory of wavelet transform and neural network and the principle of harmonic calculation in detail. Based on which the paper presents a new approach based on wavelet and neural network for the estimation of harmonic components in the power system. This proposed method preprocesses the signal using wavelet analysis to get feature extraction of signals. Then it analyses and calculates the feature vector by the artificial neural network, and the harmonic components of the current can be got. This method combines the good time-frequency localized ability in wavelet and the distributed parallel computing and adaptive learning in artificial neural network. Using the characteristics that wavelet coefficient will get the maxima at the singular point, the non-stationary signal can be detected in this method. Finally, the simulation using MATLAB for steady-state harmonic, non-stationary harmonic and nonlinear load harmonic are shown in this paper. The results show that the harmonic components can be detected at real time with high precision.