Owing to the large-scale grid connection of new energy sources, several installed power electronic devices introduce sub-/supersynchronous inter-harmonics into power signals, resulting in the frequent occurrence of subsynchronous oscillations (SSOs). The SSOs may cause significant harm to generator sets and power systems; thus, online monitoring and accurate alarms for power systems are crucial for their safe and stable operation. Phasor measurement units (PMUs) can realize the dynamic real-time monitoring of power systems. Based on PMU phasor measurements, this study proposes a method for SSO online monitoring and alarm implementation for the main station of a PMU. First, fast Fourier transform frequency spectrum analysis is performed on PMU current phasor amplitude data to obtain subsynchronous frequency components. Second, the support vector machine learning algorithm is trained to obtain the amplitude threshold and subsequently filter out safe components and retain harmful ones. Finally, the adaptive duration threshold is determined according to frequency susceptibility, amplitude attenuation, and energy accumulation to decide whether to transmit an alarm signal. Experiments based on field data verify the effectiveness of the proposed method.