Transformer Dissolved Gas Analysis Using Composite Duval Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current fault analysis methods for oil-filled transformers, such as the Duval Pentagon classification, are time-consuming and require complex geometrical calculations, making them inefficient for fleet-wide condition assessment and online monitoring of multiple transformers.
Innovation Solution
A composite fault region map is created by combining two Duval Pentagons, and a machine learning classification technique is trained to classify faults based on dissolved gas concentrations, eliminating the need for repeated geometrical calculations and enabling faster fault identification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional Duval Pentagon classification methods are used for fault analysis, then fault classification can be performed, but the process is time-consuming and requires complex geometrical calculations
Solution Approach 1:
The patent pre-divides the Duval Pentagon into multiple fault regions with predefined characteristics before actual fault analysis. By preparing the classification framework in advance with predetermined regions and their associated fault types, the system eliminates the need for complex runtime geometrical calculations, thereby reducing fault analysis time while maintaining classification accuracy
Solution Approach 2:
The patent segments the continuous Duval Pentagon space into discrete fault regions (e.g., thermal regions, electrical discharge regions, corona partial discharge regions). This segmentation transforms the complex continuous classification problem into simpler discrete region identification, reducing computational complexity and analysis time
2Measurement precision
If traditional Duval Pentagon classification methods are used for fault analysis, then fault classification can be performed, but the process requires complex geometrical calculations
Solution Approach 1:
The patent performs preliminary geometric setup by pre-defining fault regions and their boundaries before actual fault analysis. By calculating and storing region characteristics in advance, the system eliminates complex geometrical calculations during runtime, reducing computational complexity while preserving classification accuracy
Solution Approach 2:
The patent creates simplified representations of fault regions within the Duval Pentagon space. Instead of performing complex geometrical operations on the original continuous space, the system uses pre-established region models that capture essential fault characteristics, thereby simplifying calculations
3Adaptability or versatility
If fleet-wide condition assessment and online monitoring of multiple transformers are implemented, then comprehensive monitoring coverage is achieved, but the computational burden and time requirements increase significantly
Solution Approach 1:
The patent segments the monitoring task by dividing the fleet of transformers into manageable groups and further segmenting the fault analysis into discrete region identification steps. This segmentation enables parallel processing and efficient handling of multiple transformers simultaneously, maintaining high monitoring coverage while improving analysis efficiency
Solution Approach 2:
The patent prepares classification frameworks and fault region definitions in advance for all transformers in the fleet. By having predetermined region characteristics and classification rules ready before monitoring begins, the system can rapidly assess multiple transformers without repeated complex calculations, thereby improving productivity across the entire fleet
Data Source
AI summary
A method of analyzing dissolved gas in an oil-filled transformer includes determining a centroid of a polygon that represents a plurality of dissolved gas concentrations. A fault region in which the centroid of the polygon is located is determined, where the plurality of fault regions are defined in a composite fault region map that is a composite of a Duval Pentagons 1 and 2. The method classifies a fault experienced by the transformer based on the determined fault region within the composite fault region map. The classification is done by a machine learning classification technique. Further embodiments classify faults based on dissolved gas levels without determining a centroid of a polygon representing the dissolved gas levels. Related systems are also disclosed.