Transformer State Diagnosis Using Machine Learning Anomaly Plane
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current transformer diagnosis methods, particularly relying on dissolved gas analysis, struggle to accurately determine the state of transformers due to oversimplification in diagnosing anomalies, leading to frequent unexpected facility accidents and economic losses.
Innovation Solution
A method utilizing machine learning, specifically deep learning, to determine anomaly features and generate an anomaly plane, allowing for precise classification of transformer states based on decision boundaries, enabling more accurate diagnosis of transformer conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If dissolved gas analysis is used to diagnose transformer state, then the diagnostic process is simplified, but the accuracy of determining transformer state deteriorates
Solution Approach 1:
The patent transforms the diagnostic approach by changing from simple gas concentration threshold comparison to a multi-parameter machine learning model that considers multiple gas components (H2, CH4, C2H6, C2H4, C2H2, CO, CO2) and their relationships. The system extracts features from dissolved gas data and uses trained models to determine transformer states, achieving both operational simplicity and high diagnostic accuracy through automated complex analysis.
2Ease of manufacture
If simple criterion-based diagnosis is used, then the diagnostic method is easy to implement, but the ability to accurately diagnose transformer conditions deteriorates
Solution Approach 1:
The patent segments the diagnostic process into distinct modules: data collection from dissolved gas analysis, feature extraction from multiple gas parameters, model selection based on different diagnostic needs, and state determination. This segmentation allows the complex diagnostic task to be implemented systematically while maintaining high reliability through specialized processing at each stage.
Solution Approach 2:
The patent introduces machine learning models as intermediaries between the raw dissolved gas data and the final transformer state determination. These models act as mediators that process complex gas concentration patterns and relationships, translating them into reliable diagnostic conclusions without requiring direct complex rule-based logic.
3Measurement precision
If machine learning with anomaly plane is used, then the accuracy of determining transformer state improves, but the complexity of the diagnostic system increases
Solution Approach 1:
The patent develops universal machine learning models that can handle multiple transformer diagnostic scenarios and gas analysis patterns through a single anomaly plane framework. The trained models serve multiple functions including anomaly detection, state classification, and trend analysis, reducing the need for separate specialized systems while maintaining high accuracy across different diagnostic situations.
Data Source
AI summary
A method for determining an integrity factor through machine learning, and a device for performing such a method can include the steps in which: a device status determination device determines an integrity feature on the basis of machine learning; the device state determination device generates an integrity plane on the basis of the integrity feature; the device state determination device receives target input data of a target device; the device state determination device determines an integrity factor corresponding to the target input data on the basis of the integrity plane; and the device status determination device determines the status of the target device on the basis of the integrity factor.