Transformer fault on-line monitoring method and device and medium

By combining the improved CNN-RP-LSTM model with multi-sensor data collection, the problem of untimely and inaccurate measurement results in online transformer monitoring is solved, and the transformer fault can be quickly and accurately judged and located, which improves the performance and efficiency of the monitoring equipment.

CN120761731APending Publication Date: 2025-10-10SHANDONG ELECTRIC GRP DIGITAL TECH CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202510855551.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-06-25
Publication Date
2025-10-10

AI Technical Summary

Technical Problem

Among existing transformer online monitoring technologies, oil chromatography monitoring suffers from untimely and inaccurate measurement results, and the measurement accuracy of the partial discharge monitoring system needs to be improved, making it difficult to achieve fast and accurate fault diagnosis and location.

Method used

An improved CNN-RP-LSTM model based on residual network is adopted, combined with image sensor, voiceprint sensor and ultrasonic sensor to collect data. The features are extracted by the improved CNN model and input into the LSTM model for fault prediction. The reward and punishment mechanism optimization model is introduced to construct an online monitoring device for transformer faults.

Benefits of technology

It improves the real-time and accuracy of data from transformer online monitoring equipment, enhances fault judgment and location capabilities, reduces maintenance costs, improves fault response efficiency, and enhances the adaptability and interpretability of the model.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120761731A_ABST
    Figure CN120761731A_ABST
Patent Text Reader

Abstract

The invention relates to the field of power system protection and control, in particular to a transformer fault on-line monitoring method and device and a medium, and the method comprises the steps: collecting transformer data through an image sensor, a voiceprint sensor and an ultrasonic sensor, and inputting the preprocessed data into a transformer fault prediction model through a preprocessing step. A transformer fault prediction model is a CNN-RP-LSTM model, the use of an improved CNN convolutional layer enhances the capture ability of the model for data features, multiple recognition models are constructed for sampling data types, especially when image data or time series data are processed, deeper features can be extracted, and by integrating an RP award punishment mechanism and an LSTM network, the fault prediction accuracy of the transformer is improved. And the robustness of the model can be enhanced by rewarding positive behaviors and punishing negative behaviors. According to the invention, the capability of capturing data features can be improved, and the fault judgment and fault positioning capability of the transformer on-line monitoring equipment can be improved.
Need to check novelty before this filing date? Find Prior Art