Active Distribution Network Fault Detection With Graph Attention
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fault diagnosis methods for active distribution networks struggle with complex data processing, high costs, and low accuracy due to reliance on single models, poor adaptability to changing conditions, and neglect of temporal and spatial features, leading to inefficient fault detection and classification.
Innovation Solution
A data mining-enhancement framework using dynamic clustering and data enhancement, combined with a dynamic graph attention network and domain adaptation, to accurately position and classify abnormal states in active distribution networks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional fault diagnosis methods are used, then the system is simple to implement, but the diagnosis accuracy is low and cannot handle complex data effectively
Solution Approach 1:
The patent combines multiple AI algorithms (expert system, neural network, Bayesian network, fuzzy logic, Petri net, and optimization algorithms) into a hybrid fault diagnosis system. This merging of different diagnostic approaches allows the system to leverage the strengths of each method while compensating for their individual weaknesses, thereby achieving high diagnosis accuracy for complex distribution network faults.
Solution Approach 2:
The patent creates a composite diagnostic system by integrating six different AI methodologies into a unified framework. This composite approach is analogous to composite materials in engineering, where combining different substances creates a material with superior properties. The hybrid system achieves robust fault diagnosis capability that none of the individual algorithms could achieve alone.
2Adaptability or versatility
If single AI model is used, then the system is easy to operate, but the adaptability to changing network conditions is poor
Solution Approach 1:
The patent implements a dynamic hybrid AI system where the six different algorithms can be selectively activated based on the specific fault conditions and network state. The system dynamically adjusts which diagnostic methods are applied, allowing it to adapt to changing network conditions and fault types while maintaining manageable operational complexity through automated selection mechanisms.
Solution Approach 2:
The hybrid system serves multiple diagnostic functions simultaneously by incorporating six different AI methodologies, each suited for specific types of fault analysis. This multi-functionality allows the system to handle various fault scenarios (simple and complex, steady-state and transient) with a single unified platform, achieving versatility without requiring multiple separate systems.
3Measurement precision
If comprehensive data analysis is performed, then the diagnosis accuracy improves, but the processing time and cost increase
Solution Approach 1:
The patent segments the fault diagnosis process into multiple parallel tracks, each handled by a different AI algorithm. This segmentation allows comprehensive data analysis to be distributed across multiple specialized processors working simultaneously, reducing the overall processing time while maintaining high diagnostic accuracy through the collective output of all segments.
Solution Approach 2:
The system applies partial analysis through different algorithms based on the specific fault characteristics. Not all six algorithms need to process every fault scenario in full depth; instead, the system selectively engages appropriate algorithms to the extent necessary for accurate diagnosis, avoiding excessive processing while maintaining sufficient accuracy through targeted analysis.
Data Source
AI summary
Provided is a method for perceiving and eliminating an abnormal state of active distribution network based on data enhancement, including: acquiring, by synchrophasor measurement device, data of each node of active distribution network in target domain in real-time and transmitting to processor; inputting the acquired data into a classification model, and outputting abnormal detection and classification results in real time; and analyzing the abnormal detection and classification results, and transmitting an abnormal state eliminating instruction to a distribution terminal to eliminate the abnormal state. Wherein, hidden distribution features in node data of active distribution network are mined through dynamic clustering, a large amount of unlabeled data are clustered, a data label is generated through self-coding and label correction rule, training samples with balanced category distribution is generated through data enhancement and is used to train the classification model based on dynamic graph attention network by domain adaption method.