Abstract: The new power systems with high‑penetration renewables, extensive power electronics, and wide‑area interconnection place stringent demands on millisecond‑level transient response, high‑precision measurement, and intelligent diagnosis of complex equipment that cannot be met by conventional single-parameter, point-wise monitoring. State acquisition across the grid is rapidly evolving toward an intelligent perception paradigm that is multi‑physics, multi‑scale, multi-modal, and cooperative. This paper presents a structured review of intelligent perception applications in renewable generation units, transmission and substation primary equipment, and distribution/user‑side terminals. It also summarizes recent progress in key technologies, including novel magnetosensitive materials, fiber‑optic sensing, MEMS sensors, self‑powered sensing, as well as integrated multi‑parameter sensing and intelligent signal processing. On this basis, this paper analyzes the main challenges to large-scale deployment in new power systems, including limited robustness in harsh environments, constraints on sensor cost and power consumption, the pressure of massive data transmission and processing, and lagging cyber security and standardization. Future research directions are outlined in terms of high-reliability sensitive materials and novel sensing mechanisms, highly miniaturized and ultra-low-power sensors, multimodal fusion and edge intelligence, and the construction of industrial ecosystems. In conclusion, new power system development urgently requires the widespread deployment of highly reliable, miniaturized, self-powered sensing terminals, tightly coupled with edge computing, so as to build a safe, efficient and intelligent panoramic online sensing and measurement infrastructure for the power grid.