Research On Fault Line Selection In Distribution Network Based On The Fusion Of Parameter Optimized Variation Model Decomposition And Improved K Clustering Criterion

The distribution network is a crucial component that has a direct impact on the dependability of the energy supply for end-users.The occurrence probability of accidents in the distribution grid is greatly exceeds that in the transmission grid,and the likelihood of single-phase ground fault is at its peak.Two types of passive fault detection methods currently exist: utilizing stable-state measurements and transient fault measurements.However,the existing transient quantity fault section methods are susceptible to the effects of fault phase angle,transitional resistance,system noise,system harmonics,and sectioning criteria threshold.Therefore,a fault section method based on parameter-tuned variational mode decomposition(VMD)and improved K-cluster criteria fusion is recommended in this study.The main research work includes:(1)An evaluation of the single-phase grounding fault current characteristics of the low current grounding system during stable-state and transient faults.Application of an arc-suppressing coil changes the dimension and phase shift of the stable-state zero-sequence current,eliminating the fault feature,while the transient current is not impacted by the arc-suppression coil grounding system.Through the deduction of the composition of transient and fault current,it is concluded that there are differences in the components of fault current under different fault phase angles,transition resistances,and fault types.The VMD is added to the feature selection process of zerosequence current,and combined with the conclusion of differences in fault zero-sequence current components,the signal is decomposed and features are taken out.(2)Applying the variational mode decomposition technique to the wire selection in lowcurrent grounding systems,there are three key parameters that affect the final results and the efficiency of the decomposition,which need to be optimized and analyzed.The number of decomposition levels ensures the reliable separation of the modal components,the quadratic penalty factor affects the accuracy of signal decomposition,and the initial value of the central frequency of the modal element mainly affects the computational efficiency of VMD.This paper utilizes the spectrum and element features of zero-sequence current signals to ascertain the total quantity of decomposition levels of the VMD.The arithmetic optimization algorithm is used to find the best quadratic penalty factor,eliminate the interference of power frequency components,noise,and harmonic components,and finally determine the central frequency of the mode based on the decomposition levels and spectra of each mode to improve the efficiency of the variational mode decomposition.(3)In order to address the issue of low accuracy caused by the blind spots in selecting lines using a single criterion,this paper improves the K-center clustering algorithm by integrating complementary multiple criteria.This approach not only improves the accuracy of selecting lines but also avoids the problem of hard boundaries caused by setting the selection threshold.Comparative analysis shows that using the components decomposed by VMD and the proposed cosine similarity criterion improves the accuracy of selecting lines using a single criterion.The functional relationship between fault phase Angle and amplitude of fault DC and high-frequency components is obtained by analysis,and the line selection weight is determined based on this,and the advantages of the criteria are complementary.Finally,simulation analysis is conducted on the impact of fault selection caused by bus faults,long line high impedance grounding faults,arc grounding faults,intermittent arc grounding faults,and inverter-type distributed power access.The findings indicate that the suggested approach is suitable for the power system with distributed power access,and is not affected by failure time,fault location,or transitional resistance.Moreover,using the proposed technique to process multiple sets of fault zero-sequence current signals collected from actual substations can accurately identify faulty lines,which verifies the effectiveness of the proposed method.