DCNN Transformer Fault Diagnosis via Lattice Boltzmann Simulation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current fault diagnosis methods for power transformers rely heavily on external temperature monitoring and mathematical models, which become less effective with aged mechanical parts or impurities, and require frequent recalibration, while lacking in automatic detection and feature extraction capabilities.
Innovation Solution
A method utilizing a deep convolutional neural network (DCNN) combined with image segmentation, specifically the Lattice Boltzmann method (LBM) and level set method (LSM), to diagnose internal thermal faults in oil-immersed transformers by simulating temperature distributions and extracting fault features from infrared images, allowing for dynamic adjustment and precise fault positioning with minimal monitoring information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If external temperature monitoring and mathematical models are used for fault diagnosis, then the monitoring system is simple to implement, but the diagnosis accuracy deteriorates when mechanical parts age or impurities are present
Solution Approach 1:
The patent replaces traditional mathematical models and external temperature monitoring with a deep convolutional neural network (DCNN) system that processes infrared thermal images. This substitution enables automatic feature extraction from image data, eliminating the need for manual model derivation and recalibration when equipment conditions change, thereby maintaining high diagnosis accuracy without sacrificing implementation simplicity
Solution Approach 2:
The DCNN system performs self-learning and automatic adaptation through continuous training on infrared thermal image data. The network automatically adjusts its parameters and extracts relevant features without requiring external intervention or model recalibration, enabling it to maintain high diagnostic accuracy even when mechanical parts age or impurities are present in the transformer oil
2Device complexity
If traditional temperature monitoring methods are used, then the system structure is simple, but automatic detection capability is lost
Solution Approach 1:
The patent replaces manual temperature monitoring and model-based analysis with an automated DCNN system that processes infrared thermal images. The convolutional neural network automatically detects fault features, extracts temperature distribution patterns, and diagnoses internal thermal faults without human intervention, achieving high-level automation while maintaining a relatively simple system architecture through the use of standard deep learning frameworks
3Loss of information
If full infrared thermal image data is transmitted for analysis, then complete information is available, but data transmission amount increases
Solution Approach 1:
The patent extracts only the essential fault-related features from the full infrared thermal image data using the DCNN system. The network identifies and extracts key temperature distribution patterns, hotspot locations, and anomaly characteristics, transmitting only this extracted feature data rather than the complete raw image data. This extraction process preserves all necessary diagnostic information while significantly reducing the data transmission burden
Solution Approach 2:
The DCNN feature extraction module acts as an intermediary between the infrared thermal imaging system and the fault diagnosis system. It processes the full image data locally, extracts relevant features, and transmits only the essential extracted information to the diagnosis module, thereby reducing data transmission requirements while maintaining complete diagnostic capability
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enables intelligent fault feature extraction and accurate internal fault localization in power transformers with reduced data transmission and no need for frequent model recalibration, improving positioning accuracy and adaptability to changing conditions.
Implementation Method 1
Lattice Boltzmann method (LBM) has the advantages of simplicity, high calculation efficiency, and parallel processing. LBM can quickly simulate the temperature distribution inside the transformer
Implementation Method 2
Using deep convolutional neural network (DCNN) can automatically extract fault features
Implementation Method 3
image segmentation methods can be used to extract edge features and compress data
Data Source
AI summary
The disclosure provides an internal thermal fault diagnosing method for an oil-immersed transformer based on DCNN and image segmentation, including: 1) dividing an internal area of a transformer, and using fault areas and normal status as labels of DCNN; 2) through lattice Boltzmann simulation, randomly obtaining multiple feature images of the internal temperature field distribution of the transformer under normal and various fault state modes, and the fault area serves as a label to form the underlying training sample set; 3) obtaining historical monitoring information of the infrared camera or temperature sensor, and forming its corresponding fault diagnosis results into labels; 4) combining all monitoring information contained in each sample into one image, and then extracting the same monitoring information from the samples in the sample set to form a new image; 5) segmenting image sample and then inputting the same into DCNN for training to obtain diagnosis results.