Power transformer free gas detection method and system for material screening based on machine learning

By using machine learning to screen for the optimal gas-sensitive materials and constructing a sensor array, the problems of poor gas selectivity and insufficient sensitivity in the detection of ionized gases in power transformers were solved, achieving efficient and accurate online detection.

CN121601098AActive Publication Date: 2026-03-03CHONGQING UNIV +1
5 Cites 4 Cited by

Patent Information

Application Number
CN202610128935.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-01-30
Publication Date
2026-03-03
Estimated Expiration
2046-01-30

AI Technical Summary

Technical Problem

Existing technologies for detecting free gas in power transformers suffer from poor gas selectivity, insufficient material sensitivity, and poor environmental adaptability. Traditional methods cannot achieve high sensitivity, high selectivity, and high stability in online monitoring.

Method used

A machine learning-based material screening method was adopted, which used high-throughput first-principles calculations and CGCNN convolutional kernels fused with the GNN framework to screen the optimal gas-sensitive materials. Then, a sensor array was constructed for detection by combining a Bayesian optimization framework and data analysis algorithms.

Benefits of technology

It enables rapid, accurate, and online detection of various key free gases in transformer oil, improving material development efficiency and sensor performance. It can identify a variety of characteristic gases with high sensitivity, overcoming the limitations of traditional sensors.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention provides a power transformer free gas detection method and system for material screening based on machine learning, and the method comprises the steps: carrying out the sampling of a compound space, and calculating a material characteristic data set of different doping configurations for the adsorption performance of a target gas through a high-throughput first principle; based on the material characteristic data set, a CGCNN convolution kernel is fused through a GNN framework, and a pre-trained material screening model is obtained and used for predicting the adsorption performance of different doping configurations to target gas; screening by utilizing a pre-trained material screening model to obtain an optimal doping configuration meeting a preset performance threshold value; screening based on the optimal doping configuration to obtain an optimal gas sensitive material; acquiring a response signal of the sensor array to the dissolved gas removed from the transformer oil, identifying the type of the target gas, and estimating the corresponding gas concentration; wherein the sensor array is prepared from an optimal gas sensitive material. According to the invention, rapid, accurate and on-line detection of various key free gases in transformer oil can be realized.
Need to check novelty before this filing date? Find Prior Art