Research On Passive Islanding Detection Method For Grid Connected Distributed Generation Based On Wavelet Transform And BP Neural Network

With the increasing scale of the distributed generation system,the impact of the distributed generation system on the original power supply network is bigger and bigger,and multiple factors need to be considered.Among them,islanding detection is an important problem that must be considered for the stable operation of the distributed generation system.The existing active islanding detection technology reduces the power quality due to the disturbance of the injected signal to the power grid.At the same time,due to the dilution effect of the injected disturbance signal in the case of multiple inverters,it is difficult to accurately measure the islanding state,so researchers turn their attention to the passive islanding detection method.Traditional passive islanding detection technology needs time to detect the characteristic signal to make a judgement,which results in slow detection speed,large non-detection zone and unable to identify islanding state under power matching.In view of the problems existing in the traditional passive islanding detection method,this paper analyzes the mechanism of the islanding effect,introduces the technology of wavelet transform and neural network,and proposes a passive islanding detection algorithm based on wavelet transform and BP neural network.Using the data mining characteristics of wavelet and the powerful pattern recognition ability of neural network,it realizes faster detection speed and smaller non-detection zone.The main research work and innovation are as follows:(1)In this paper,the mechanism of the islanding effect is derived in detail,and the influence of active power and reactive power mismatch on the voltage amplitude and frequency of the PCC point is analyzed.Then,the principle and existing problems of traditional passive islanding detection methods of over/under voltage and over/under frequency are analyzed,and the load quality factor in the anti-islanding test circuit is explained.(2)Given of the problem that it is difficult to select the eigenvector of islanding detection,based on introducing the theory of wavelet transform,according to the general selection principle of wavelet basis function and the optimal wavelet basis selection method of transient non-stationary signal,the optimal wavelet basis is obtained,and the eigenvector of islanding detection in this paper is constructed,and the effectiveness of the selected eigenvector is verified by simulation.(3)To solve the problem of islanding recognition when the power required by local load matches the output power of inverter,the pattern recognition and classification technology of BP neural network is introduced,and the BP neural network model for the anti-islanding test is constructed according to the selected islanding signal eigenvector,and the key parameters such as the number of nodes,hidden layers and activation function parameters of the neural network is determined in turn.(4)Aiming at the problem that the existing passive island detection method is difficult to detect islanding state during power matching,two passive islanding detection algorithms based on wavelet transform and BP neural network are proposed,and a 10 k W level distributed generation multi inverter islanding detection algorithm verification platform is built to verify the effectiveness of the proposed islanding detection algorithm under various load parameters and non-islanding interference conditions.