Machine Learning-Driven Framework for Reducing PAPR in Satellite Communication Systems

High peak-to-average power ratio (PAPR) in orthogonal frequency division multiplexing (OFDM) signals presents a persistent challenge in satellite communications (SatCom), impacting signal quality and causing adjacent channel interference. This paper introduces a novel framework that combines the elastic net-based machine learning (ML) model with the partial transmit sequence (PTS) technique to effectively reduce PAPR. Additionally, the potential of artificial intelligence (AI) approaches are investigated, specifically swarm intelligence and ML methods, for high-performance, low-complexity solutions. In this regard, ML models are applied to mitigate PAPR in SatCom networks under the presence of a traveling wave tube amplifier (TWTA) model and a land mobile satellite (LMS) channel, employing 16-quadrature amplitude modulation (16-QAM). Compared with the baseline schemes, simulation results demonstrate that the proposed ML framework, integrating principal component analysis (PCA) with the elastic net learning model, achieves comparable PAPR performance and minimal computational complexity.