Research on the Bird Activity Monitoring Method for Overhead Transmission Lines Based on DAS
Year: 2026
Type: article
Abstract: To address the challenge of full-time and all-space monitoring of bird activity along long-distance transmission corridors, this article proposes a bird activity monitoring method integrating distributed acoustic sensing (DAS) and frequency-domain-driven deep learning. First, optical fiber composite overhead ground wire (OPGW) was employed to acquire acoustic signals along the transmission line. Adaptive extraction of bird vocalization frames was achieved using a dual-threshold constraint combining subband energy and spectral kurtosis. Variational mode decomposition (VMD) and permutation entropy (PE) were further applied to suppress environmental noise. Subsequently, a bird species identification network incorporating frequency-domain attention and multiscale receptive fields was constructed. A multiscale selective convolution architecture captured fine- and coarse-grained vocal features of bird sounds, while a discrete cosine transform (DCT) attention mechanism was introduced to effectively enhance the model’s capability in extracting acoustic features. An outdoor transmission-line test platform equipped with OPGW was established to build an eight-species acoustic database of birds posing high risks to transmission line faults, enabling validation of the proposed model’s accuracy. Experimental results demonstrate that the frequency-domain-driven multiscale convolutional network (FDSKNet) model achieves an average accuracy of 95.99% and maintains an accuracy of 80.75% under a harsh noise condition of 0-dB signal-to-noise ratio, with overall performance superior to other classical algorithms, providing a distributed technical solution for bird activity monitoring along transmission lines.
Source: IEEE Sensors Journal
Institution: North China Electric Power University
Language: English
Subfield: Ecology
Field: Environmental Science
Domain: Physical Sciences
Open Access status: hybrid
Estimated APC paid: $2,645
OpenAlex ID: W7202127639