Distribution Network Fault Detection Based On Wavelet Transform And Singular Value Decomposition

The distribution network is in the terminal of the power network, which is responsible for the distribution of electric energy, and the security and reliability of the operation will directly affect the quality of power supply. When the short circuit fault occurs in distribution network,if the fault can not be detected in time and quickly removed, it will cause damage to the power equipment, and even affect the stability of the whole power system. Therefore, the fault of distribution network to achieve rapid and accurate detection and classification of fault types, not only can improve the stability and quality of power supply of power system, but also can reduce economic losses caused by the power outage, and maintain social stability and order.Aiming at the destructiveness of short circuit fault in distribution network, in this thesis,the theory of wavelet transform and singular value decomposition are combined,and a new method of fault detection and classification for distribution network is proposed based on the maximum wavelet singular value. Firstly, each phase current signal is collected from the first end of the distribution network,and then the current signal is continuously transformed by a series of high pass and low pass filters, obtaining the wavelet detail coefficients of the current signal in each frequency band. Selecting the fifth layer wavelet detail coefficients as the characteristic sequence of distribution network fault detection in this paper and selecting a half AC cycle as the fault detection time window length, the fifth layer wavelet detail coefficients reconstructed the feature matrix reflected the fault characteristics of distribution line. In the end, singular value decomposition is performed to obtain the maximum wavelet singular value of each time window.When a short circuit fault occurs in the distribution line, the maximum wavelet singular value will change suddenly in a half AC cycle mutation,and its sensitivity is high. According to the mutation degree of the maximum wavelet singular value, the fault occurrence time can be detected. When short circuit fault occurs,the maximum wavelet singular value will show a certain change rule,according to the change rule,single-phase grounding short circuit, two-phase grounding short circuit, two-phase short circuit and three-phase short circuit fault are divided. In the actual operation of distribution network, not only the metal short circuit fault occurs, but also the high resistance short circuit fault occurs. In view of the short circuit fault under different impedance, this paper used MATLAB to build a typical distribution network model,and analyzed the short circuit fault under different impedance and different short circuit point. Results show that the proposed fault detection and classification method based on maximum wavelet singular value can maintain the accuracy and sensitivity of detection and classification in the short circuit impedance and has a strong practical value.