This paper presents a real-time structural health monitoring (SHM) framework for wind turbine towers using deep learning models applied to vibration data collected from an IoT-based wireless sensing system. The proposed system integrates high-resolution triaxial accelerometers (ADXL345), GPS-enabled ESP32 microcontrollers, and 4G communication for distributed deployment on turbine towers. Raw vibration data is preprocessed using FFT and statistical filtering before being transmitted to a cloud server via MQTT. A deep autoencoder-based anomaly detection model is trained on normal operating vibration patterns to identify structural deviations that may indicate early-stage damage or fatigue. The system is capable of real-time inference and alarm generation, supporting preventive maintenance strategies in remote wind farm environments. Field experiments on scaled turbine structures demonstrate that the proposed method detects subtle frequency shifts and amplitude anomalies with high sensitivity and low latency. This work contributes a cost-effective and scalable framework for applying AI in the structural monitoring of renewable energy infrastructure.