Underground power cables (UGCs) are very important for modern distribution networks, but they are still prone to problems like partial discharges, thermal hotspots, and insulation degradation. Traditional diagnostic techniques are constrained by elevated costs, susceptibility to noise, and insufficient real-time adaptability. This paper proposes an Internet of Things (IoT)-enabled framework for the detection and localization of underground cable faults, integrating intelligent sensors, edge analytics, and secure communication protocols. A distributed sensor suite that includes high-frequency current transformers (HFCT), acoustic emission sensors, optical fiber interferometry, and environmental probes records important fault signatures. A lightweight quantized CNN-GRU model extracts features on-node and sorts them, allowing for real-time inference with little energy use. Tests in the lab and in the field show that it works better, with a detection F1-score of over 0.92 and a localization mean absolute error of less than 1.8% of cable length. Energy-latency trade-offs prove that battery-powered IoT nodes are a good fit. Blockchain-inspired methods make it possible to share data safely. The results show that IoT-driven methods can greatly improve predictive maintenance, shorten the time it takes for outages to happen, and raise utility reliability indices.