Method and system for positioning fault of power distribution network

Abstract

The invention discloses a method for positioning the fault of a power distribution network. The method comprises the following steps: carrying out machine learning training on a deep neural network model frame containing a multilayer network module and a bidirectional long and short time memory network module, and obtaining an optimal deep neural network model; carrying out working condition waverecording on the power distribution network in each monitoring terminal to obtain wave recording data, and intercepting the wave recording data to obtain a fault waveform area; using the multilayer network module in the optimal deep neural network model to carry out feature extraction on the fault waveform area; uploading feature data to a system master station in each monitoring terminal, and carrying out feature data collection on the system master station, and combining the feature data of the monitoring terminals on the same transmission line into a feature data sequence according to a topological structure of the power distribution network; and inputting the feature data sequence into the bidirectional long and short time memory network module so as to obtain relative positions between each monitoring terminal and the fault point.

Description

technical field [0001] The invention relates to the technical field of electric power detection, in particular to a method for locating a distribution network fault. Background technique [0002] CN103728532 discloses a method for installing a special distribution automation feeder terminal in a distribution line sectional switch to locate a ground fault. In this invention, the distribution automation feeder terminal collects zero-sequence voltage 3U0 and zero-sequence current 3I0, and then performs a series of processing on 3U0 and 3I0 to extract features, and then uses the preset fault judgment rules to determine where the current distribution automation feeder terminal is located. The position of the switch relative to the ground fault point is finally combined with multiple distribution automation feeder terminals to locate the fault section. [0003] In the above technical solution, only the fault information collected by a single terminal itself is used for fault judg...

AI Technical Summary

Problems solved by technology

[0005] However, in order to capture enough ground fault information, the fault recording will adopt a higher sampling rate, usually a single recording usually reaches tens of Kbytes in size, resulting in the following problems: 1. Recording transmission takes a long time , has a certain impact on the real-time performance of fault judgment; 2. In the season of high fault occurrence, multiple line faults may occur within a short period of time in the same main station jurisdiction, because the collection of a single wave recording often goes through dozens of times The data interaction may lead to a large amount of data flow into the main station within a period of time, so the main station needs to have the ability to process a large amount of data traffic for a long time, and the performance requirements of the main station are relatively high; 3. Most of the current The power distribution terminal for fault recording uses wireless network to communicate with the main station, a large amount of recording data will generate a lot of wireless charges, and the operating cost is relatively high

[0006] At the same time, the fault location method in the prior art divides the waveform feature extraction and fault location judgment into two steps, that is, the multi-position original waveform in the distribution network topology obtained by the terminal equipment in the distribution network needs to be manually extracted first. feature, and then use the feature to locate the fault. This fault location judgment scheme does not use the end-to-end training method of direct feedback, so the accuracy of fault location recognition is limited.

Images