Method Research And Application On Single-phase Grounding Fault Location Of Distribution Network Based On Learning Vector Quantization Neural Network

In the low-voltage distribution network of our country, about 80% of the grounding fault is single-phase grounding fault. when the single-phase grounding fault occurs, the low impedance short circuit can’t form in the system, as a result, the fault current is much smaller than the load current, which makes it difficult to find out the fault line and the fault position quickly and accurately. If the fault can’t be ruled out and the power supply can’t be restored in time, it will cause significant economic losses. Therefore, it is a significant importance to ensure the safety and reliability of the network’s running, and detects the fault line and position quickly and efficiently.Although the study of fault line selection and fault location in small current grounding system has been carried on by the Domestic and foreign researchers for a long time, and various methods of fault line selection and fault location have been proposed, the results are not ideal, the accuracy of fault line selection and fault location device has been low. Based on the study of domestic and foreign scholars on the issue of small current grounding fault, this paper uses the transient and steady-state characteristics of the fault signal to mainly study the following contents.This paper studies a method of fault line selection based on LVQ neural network.In the experiment, we change the fault position, fault phase and the size of grounding resistance to obtain a lot of zero-sequence current signals in different grounding condition, then we can obtain the fault characteristics(the active power, the wavelet energy and the fifth harmonic component) after the analysis. Finally, input these characteristics into the MATLAB simulation model of LVQ to get the results of line selection.This paper designs a remote monitoring system interface of the distribution network, which makes it more convenient for the users to view the information of fault diagnosis.On the basis of fault line selection, this paper proposes an improved algorithm of LVQ neural networks for fault location. Although the LVQ algorithm has an excellent performance in line selection, it has various shortcomings in fault location. Theimproved algorithm for fault location is based on the LVQ, there are some improvements in the process of the training, the initial value and other aspects. After the simulation experiment, the results shows that the improved algorithm has a much higher convergence speed and accuracy of fault location than before.At the same time, this paper introduces the other two fault location algorithms which are widely used currently---the wavelet neural network location algorithm and the location algorithm based on internal model, then we take a detailed comparison of the advantages and disadvantages between the improved LVQ algorithm and these two algorithms.Finally,this paper introduces the application design of fault location system, and tests the feasibility of the algorithms and the function of the system on the testing platform in the lab.