Intelligent Monitoring of Powerline Vibrations - Sparse Sensing and Predictive Modeling Approach

Abstract

Wind-induced vibrations (WIV), particularly Aeolian vibrations, pose a persistent threat to the structural integrity and operational lifespan of overhead power transmission lines. Traditional mitigation strategies, such as passive dampers, often lack adaptability to changing environmental conditions and fail to provide real-time insights into the dynamic behavior of conductors. This thesis presents a data-driven framework for real-time monitoring and state estimation of vibration profiles in transmission lines, offering a scalable and intelligent alternative to conventional methods.The proposed approach integrates a range of techniques including LSTM-based neural networks for time-series prediction, Dynamic Mode Decomposition (DMD) with Tikhonov regularization for stable reduced-order modeling, and estimation algorithms such as Kalman and particle filters for reconstructing full-system states from sparse observations. A pivoted QR decomposition method is employed to determine optimal sensor placement, ensuring efficient coverage and accurate reconstruction with minimal hardware. Additionally, compressive sensing is introduced to recover missing data, enhancing system robustness under conditions of partial sensor failure or data loss. Simulation results validate the effectiveness of the proposed methods, demonstrating accurate vibration profile reconstruction and reliable antinode tracking with a reduced sensor set. The framework provides a foundation for intelligent asset management in power systems, with implications for vibration mitigation, infrastructure resilience, and smart grid applications.