Research On Small Current Grounding Fault Location Method Based On Multi-feature Fusion

The rapid fault location,fault isolation and fault removal of distribution line is important for ensuring the safety and stable operation of the entire power system.Among them,when the single-phase grounding fault occurs in the small current grounding system,the steady-state signal is weak and the single selection criterion is easily affected by the fault conditions,the actual application effect of the current fault detection methods is not ideal.Therefore,this paper concerns the problem of section location in fault location of small current grounding system and focuses on solving the problem of misjudgment of fault section under limit fault conditions.The main work of this paper is as follows:(1)The theoretical model of single-phase grounding fault in small current grounding system is established,and the steady-state and transient characteristics of single-phase grounding fault are deduced in detail.The zero-sequence current under different fault conditions is comprehensively simulated and verified by ATP/EMTP simulation software,which verifies the reliability of theoretical analysis and lays a theoretical foundation for the subsequent selection of single characteristic quantity.(2)The feasibility of variational mode decomposition(VMD)in processing the zero-sequence current signal is analyzed in detail,and the selection of the optimal value of the key parameters of the algorithm is discussed in combination with the characteristics of the signal itself.For the number of decomposition layers,the optimal value is determined by using the principle of maximum energy of the main resonance component o the zero-sequence current when the single-phase ground fault occurs.For the penalty factor,determine its optimal value by comparing whether there are false components under different values.Then,the adaptive VMD is used to decompose the transient zero-sequence current collected by the feeder terminal into three intrinsic mode components(IMF).Finally,five characteristic quantities that can fully reflect the difference in zero sequence current at each detection point of the fault line in transient information are mined: the difference of the fault detection points are excavated: the speed of the decaying DC component,the ratio of the a peak value of the transient high-frequency component to the amplitude of the steady-state frequency component,the high frequency transient component,the energy ratio of the transient high frequency component and decaying DC component,the difference coefficient of fault zero-sequence current waveform between adjacent detection points,the polarity ration of transient high frequency component of adjacent detection points.(3)The feature fusion method based on machine learning is studied.Firstly,the basic principle and background of support vector machine(SVM)and traditional whale optimization algorithm(WOA)are described.Then,based on the traditional whale algorithm,the whale optimization algorithm is improved by integrating multiple strategies to solve the problem of SVM parameter optimization.Finally,the simulation model is used to simulate various fault conditions of the small current grounding system,and the fault samples are constructed.The comparison of two fault location models verifies that the fault location model optimization algorithm has a high recognition rate.Furthermore,a single feature quantity is used as the input vector of the support vector machine to construct multiple fault localization models,the test results show that the integrated segment location algorithm proposed in this paper has higher reliability.