With the implementation of the “Dual Carbon” strategy, the large-scale integration of renewable energy sources and power electronic equipment into the power grid has intensified the problem of wideband oscillations. Traditional monitoring techniques suffer from limited bandwidth and poor noise resilience, making them inadequate for the real-time and accuracy demands of modern power systems. This paper proposes a wideband oscillation monitoring approach based on a denoising autoencoder (DAE) and a gated recurrent unit (GRU) network. At the substation side, the DAE is utilized for signal compression and feature extraction, effectively overcoming communication bandwidth constraints. At the control center, the GRU is employed for oscillation detection, eliminating the need for manually defined thresholds. Upon oscillation detection, the DAE decoder reconstructs the signal to enable global analysis. Simulation results on the IEEE 39-bus system validate the proposed method’s high accuracy and robustness under various noise levels and data loss scenarios.