UHVDC transmission system oscillation source identification method based on wavelet analysis

By combining scale-channel joint attention network and bidirectional long short-term memory network, the problems of wavelet basis selection relying on experience and multi-oscillation source coupling are solved, realizing adaptive identification and dynamic tracking of oscillation sources in UHV systems, and improving identification accuracy and real-time performance.

CN121529582BActive Publication Date: 2026-05-15TIANJIN UNIV +2
2 Cites 0 Cited by

Patent Information

Application Number
CN202610051275.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2026-01-15
Publication Date
2026-05-15
Estimated Expiration
2046-01-15

AI Technical Summary

Technical Problem

In existing technologies, the selection of wavelet basis functions relies on human experience, lacks a unified standard, makes it difficult to adaptively adjust the analysis strategy, cannot effectively separate spectral aliasing in scenarios with multiple oscillation sources, and lacks the ability to track the dynamic evolution of oscillation sources, leading to misjudgment or omission of oscillation source location.

Method used

By employing a scale-channel joint attention network and a bidirectional long short-term memory network, and through adaptive weight allocation of branches of Mohr wavelet, Daubsey wavelet, Sim wavelet and Koff wavelet, combined with multidimensional electrical quantity feature analysis, the spatiotemporal evolution law of the oscillation source is extracted, thereby achieving frequency band adaptive fusion and dynamic tracking.

Benefits of technology

It improved the accuracy and real-time performance of oscillation source identification in ultra-high voltage systems, increasing the success rate from 42.3% to 78.4% and the positioning accuracy from 67.3% to 91.2%, while meeting the requirements for real-time analysis in terms of calculation time.

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

Abstract

The application relates to the technical field of power system oscillation analysis, and discloses a method for identifying an ultra-high voltage direct current sending-out system oscillation source based on wavelet analysis. The method comprises the following steps: collecting electrical quantities of measuring points, pre-processing the electrical quantities to construct an input tensor, inputting the input tensor into four wavelet branches to perform five-layer convolution pooling to obtain feature maps, adaptively weighting and fusing the feature maps through a scale channel joint attention network, inputting the feature maps into a bidirectional long short-term memory network to calculate an oscillation source probability, an energy flow direction and a frequency energy in combination with historical states, and determining an oscillation source position according to a comprehensive criterion. The application solves the problems that wavelet basis selection depends on experience, multiple oscillation sources are difficult to separate, and dynamic tracking capability is lacking, and improves the accuracy and real-time performance of ultra-high voltage system oscillation source identification.
Need to check novelty before this filing date? Find Prior Art