This work integrates wireless energy harvesting with deep learning-based spectrum prediction to propose a novel energy-efficient resource allocation strategy for Internet of Things networks supported by cognitive radio. In order to maximize energy efficiency (EE) while respecting interference limits, data rate fairness, user buffer state, and energy causality limitations, we suggest a mixed-integer nonlinear programming model. We provide an enhanced mixed integer linear programming (EMILP) model in conjunction with the distributed chicken optimization (DCO) technique to tackle the computational difficulties in resource allocation, guaranteeing near-optimal outcomes with less complexity. This study's primary novelty is the use of a long short-term memory (LSTM) neural network for channel occupancy prediction, which narrows the search space and speeds up convergence. Our comprehensive simulations show that the suggested LSTM-EMILP-DCO framework reduces computing time by 11.5% while achieving 10.8% greater EE, 9.6% better spectrum efficiency, and 8.4% higher throughput when compared to current approaches.
Pradeepa et al. (Sun,) studied this question.