<![CDATA[The growing need for sustainable agricultural practices has spurred the integration of cyber-physical systems (CPS) into modern farming. This paper presents the design, deployment, and evaluation of a modular CPS architecture for adaptive energy monitoring and control in smart agriculture. The system integrates environmental sensing, predictive modelling, and optimisation-guided actuation to enhance energy efficiency and operational resilience. Field tests on a 3-hectare site across six crop environments demonstrated significant performance gains, achieving energy savings of up to 25.8% and peak demand reductions of up to 19.8%. Our multi-layer architecture, featuring STM32 microcontrollers, LoRaWAN communication, and a cloud analytics dashboard, enables proactive control by anticipating energy demand using an LSTM-NARX predictive model. This approach reduced control actuation delay to 1.8 seconds and proved robust against cyber-physical faults, recovering from communication failures and data anomalies in under 15 seconds. The results validate that embedding energy-aware, predictive logic into CPS infrastructure creates scalable, efficient, and reliable agricultural solutions. We acknowledge limitations related to predictive model complexity and communication latency, and we propose future work focused on distributed CPS coordination, federated learning, and full lifecycle sustainability analysis to further advance intelligent, resource-efficient agriculture.]]>
