PMU advancements and applications in the AI Era

Phasor Measurement Units (PMUs) have long been central to wide-area power system monitoring. However, the growing penetration of inverter-based resources (IBRs) and emerging data-driven analytics highlight limitations of conventional phasor abstractions in capturing fast electromagnetic dynamics. This paper provides a critical review of recent PMU advancements, emphasizing the transition from phasor-based measurements toward high-fidelity point-on-wave (PoW) sensing. Unlike existing surveys, this work evaluates the state of the art across three dimensions: (i) hardware evolution toward edge-capable sensing, (ii) AI-enabled estimation and analytics for waveform-scale data, and (iii) deployment strategies in complex environments such as microgrids and data-center-rich grids. Key bottlenecks in data acquisition and processing are summarized, and representative commercial PMU solutions are compared with emerging PoW-oriented technologies from both capability and cost-implication perspectives. An illustrative field case is included to show how waveform-level processing can reveal oscillatory components that may be attenuated under conventional multi-cycle processing. The review concludes with a strategic roadmap to bridge academic AI models with practical grid applications.