In order to significantly improve the energy efficiency of elastic optical forward（EOF） networks, a functional segmentation selection and resource allocation optimization scheme in EOF is proposed. This scheme solves the complex time averaged stochastic optimization problem in functional segmentation and resource allocation by introducing improved Lyapunov drift technology, achieving dynamic and efficient allocation of network resources. In addition, the positive effects of increasing the number of antenna ports and refining the fiber spectrum granularity on the performance of the scheme were also thoroughly explored. The simulation results show that compared with the traditional cloud wireless access network based on optical fronthaul（C-RAN）, this solution not only significantly reduces average power consumption by up to 70%, but also ensures that the forward latency is maintained below 250 μs, demonstrating excellent performance advantages.