A flat reactor working condition identification method and a computer readable storage medium

By constructing an adversarial feature-enhanced neural network and utilizing adversarial game strategies and sparse feature extraction, the "domain offset" problem in the operating condition identification of smoothing reactors was solved, achieving efficient and accurate operating condition identification and ensuring the safety of the high-voltage direct current transmission system.

CN122132966APending Publication Date: 2026-06-02XI AN JIAOTONG UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
XI AN JIAOTONG UNIV
Filing Date
2026-03-19
Publication Date
2026-06-02

AI Technical Summary

Technical Problem

Existing technologies suffer from "domain offset" problems in identifying the operating conditions of smoothing reactors. The simulation data differs greatly from the field data, resulting in low identification accuracy and low training efficiency, which cannot meet the needs of real-time field diagnosis.

Method used

An adversarial feature enhancement neural network is constructed to extract and classify features from source and target domain data through adversarial game strategies. Wavelet decomposition, normalization, and slicing techniques are employed, combined with sparse feature extraction and multilayer perceptron, to optimize the loss function in order to eliminate domain differences and noise and improve recognition accuracy.

Benefits of technology

It effectively solves the "domain offset" problem, improves the accuracy of smoothing reactor condition identification, reduces the false alarm rate and the missed alarm rate, and ensures the safe and stable operation of the high voltage DC transmission system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122132966A_ABST
    Figure CN122132966A_ABST
Patent Text Reader

Abstract

The present application belongs to the technical field of electrical devices, and relates to a flat reactor working condition recognition method and a computer readable storage medium. The method comprises the following steps: constructing a source domain data set and a target domain data set; denoising and normalizing the data in the source domain data set and the target domain data set, and slicing to obtain input samples; constructing an adversarial feature enhancement neural network composed of a feature extractor, a working condition classifier and a domain discriminator; inputting the input samples into the adversarial feature enhancement neural network in sequence, training the adversarial feature enhancement neural network using an adversarial game strategy, fixing the domain discriminator parameters to obtain a trained adversarial feature enhancement neural network when the total loss function meets the optimization target; and inputting the preprocessed real-time collected flat reactor electrical signal data into the trained adversarial feature enhancement neural network to determine the current working condition category. The present application can effectively remove signal redundancy and noise, and has high working condition recognition accuracy.
Need to check novelty before this filing date? Find Prior Art