This study presents the development of a livestock behavior classification system that emphasizes energy efficiency and long-term operation by eliminating the need for conventional batteries. By integrating radio frequency energy harvesting with deep learning-based analysis, the system aims to enhance both sustainability and behavioral monitoring accuracy in smart farming environments. The proposed solution incorporates a wearable accelerometer module powered by a 915 MHz RF energy-harvesting unit, featuring an energy-optimized hardware architecture to minimize power consumption. The overall system comprises four main components: the energy-harvesting accelerometer module, an RFID reader, an Internet of Things (IoT) gateway, and a centralized data storage server. Acceleration data collected from livestock are transmitted via the RFID chip to the IoT gateway and then forwarded to a server, where deep learning algorithms are applied for behavior classification. Under optimal RF input conditions, the system achieves a power conversion efficiency of up to 80% and a behavior classification accuracy of 95%. Real-time data communication is maintained with minimal energy usage, ensuring consistent and reliable operation. Furthermore, by replacing conventional supercapacitors with the SLB08115L140 rechargeable lithium–titanate battery, the system benefits from enhanced energy retention, enabling maintenance-free livestock monitoring. This integration of RF energy harvesting, advanced energy storage, wearable sensing, and AI-driven analysis provides a self-sustaining, high-performance solution for next-generation smart livestock management.