Power Distribution Fault Location Using Mixed-Mode Wave Recording
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fault location methods in power distribution networks require manual extraction of waveform features, which is inefficient and limits the accuracy of fault location, especially for intermittent faults.
Innovation Solution
A multi-channel deep neural network is constructed to process high-frequency and low-frequency data blocks from waveforms, using multi-layer network modules and LSTM cells to identify fault locations directly from mixed recording waves.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual extraction of waveform features is used, then the process is simple and easy to implement, but the accuracy of fault location is limited and efficiency is low
Solution Approach 1:
The patent replaces the manual mechanical feature extraction process with an automated deep neural network system. The multi-channel deep neural network automatically processes mixed mode wave recording data to identify fault locations, eliminating the need for manual waveform feature extraction and achieving both high accuracy and efficiency through automated end-to-end processing
Solution Approach 2:
The deep neural network system performs self-service by automatically extracting features and determining fault locations without human intervention. The system uses multi-channel processing to simultaneously analyze multiple waveform channels, automatically identifying fault characteristics and locations through the trained neural network model
2Loss of information
If existing machine learning models are applied with truncated wavelet segments, then the model processing capacity is manageable, but valuable information from long-span waveforms is lost
Solution Approach 1:
The patent applies segmentation by dividing the complex processing task into multiple parallel channels that process different waveform segments simultaneously. The multi-channel deep neural network processes different time segments and frequency components in parallel, then integrates the results to achieve comprehensive fault detection without losing valuable long-span waveform information
Solution Approach 2:
The patent transforms the problem from processing truncated 1D wavelet segments to processing multi-channel mixed mode wave recording data in the time domain. This dimensional transformation allows the system to utilize long-span waveform information across multiple channels simultaneously, preserving valuable temporal and spectral information while managing computational complexity through efficient multi-channel processing
3Extent of automation
If two-step process (feature extraction then fault location) is used, then the process is straightforward and easy to implement, but end-to-end fault location identification cannot be achieved
Solution Approach 1:
The patent merges the feature extraction and fault location determination steps into a single integrated deep neural network system. The multi-channel network simultaneously performs feature extraction from mixed mode wave recording and fault location identification in one end-to-end process, achieving full automation while managing complexity through unified model architecture and training
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present disclosure relates to the field of power technology, and in particular to a system for locating fault in power distribution network based on mixed mode wave recording.