A power transformation equipment sound imaging monitoring method and system based on a physical embedding neural network

By using a physical embedded neural network approach, a circular array monitoring system, and multi-iteration inversion technology, the problems of low acoustic imaging resolution and long computation time of power equipment were solved, achieving efficient and accurate fault identification and judgment.

CN122241518APending Publication Date: 2026-06-19BINZHOU POWER SUPPLY COMPANY OF STATE GRID SHANDONG ELECTRIC POWER
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
BINZHOU POWER SUPPLY COMPANY OF STATE GRID SHANDONG ELECTRIC POWER
Filing Date
2026-03-19
Publication Date
2026-06-19

AI Technical Summary

Technical Problem

Existing acoustic imaging monitoring technologies for power equipment suffer from problems such as low imaging resolution, ambiguous fault location, insufficient reliance on prior information, and long computation time, making it difficult to meet the needs of rapid monitoring.

Method used

A method based on physical embedded neural networks is adopted. Time-domain sound field data is collected through a circular array monitoring system to construct a real frequency-domain sound field feature matrix. The physical embedded neural network PEN-FWI is used to perform multi-iteration inversion correction of the acoustic model, and the structural similarity index SSIM is combined for fault determination.

Benefits of technology

It achieves high-precision and rapid acoustic imaging of power equipment, which can clearly identify minor fault areas, reduce reliance on prior information, improve the universality and computational efficiency of monitoring, and reduce human interpretation errors.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122241518A_ABST
    Figure CN122241518A_ABST
Patent Text Reader

Abstract

This invention belongs to the field of power equipment technology, specifically relating to a method and system for acoustic imaging monitoring of substation equipment based on a physically embedded neural network. It includes a circular array monitoring system arranged around the substation to be monitored, utilizing several transceiver-integrated acoustic sensors to collect time-domain sound field data; and employing a uniform underwater acoustic model as the initial acoustic model for multi-iterative inversion. This eliminates the need for prior knowledge of specific acoustic information such as the casing material, internal structural parameters, and component acoustic impedance distribution of the substation under test. High-precision imaging can be achieved simply by training the physically embedded neural network using a general substation acoustic model sample. This overcomes the strong dependence of traditional FWI technology on prior equipment information, avoiding imaging failures caused by getting stuck in local optima due to a lack of prior information. Furthermore, it eliminates the need to adjust monitoring parameters for different substations, significantly improving the technology's versatility and lowering the operational threshold for on-site monitoring.
Need to check novelty before this filing date? Find Prior Art