Traditional hydrogels tend to freeze and lose performance at low temperatures, limiting their applications. Additionally, hydrogels need to exhibit low hysteresis, excellent cycling stability, and self-adhesion to ensure high-quality signal acquisition in complex environments. To address this challenge, this study designed a dual-network gel in a glycerol (Gly)/H2O solvent system. Due to the combination of chemical and physical crosslinking (hydrogen bonding and electrostatic interactions), the resulting gel exhibits skin-adaptive modulus, high cycling stability, anti-freezing ability, body temperature-induced adhesion, and excellent electrical performance, making it suitable for wearable sensors at low temperatures. Based on this gel, a single-electrode triboelectric nanogenerator (gel-TENG) is developed, achieving efficient conversion of mechanical energy into electrical energy. Further applied to a smart insole, it successfully enabled real-time visualization of plantar pressure distribution and skiing motion recognition. Using a random forest machine learning algorithm, the system accurately classified 11 basic skiing motions, achieving a classification accuracy of 97.1%. This study advances flexible sensors and self-powered systems, supporting intelligent materials research in extreme environments.