Power optimization in 5G network bands using modified genetic algorithm

Abstract

This study investigates power optimization techniques in fifth-generation (5G) networks to enhance network performance across multiple metrics. Slow con vergence and often getting stuck in local optima is an issue in traditional ge netic algorithm (GA). Whereas, with machine learning (ML) algorithms large datasets are required and constraints cannot be enforced. To avoid local min ima and achieve global optimal solution, the average fitness value is checked in modified GA after completion of the total count Nc, so that the change in fitness should be greater than 1%. In this research work, modified GA algorithm is used to optimize power allocation, minimize interference, and improve resource man agement across different frequency bands, specifically 700 and 3500 MHz. The results show that the modified GA achieves enhancement in spectral efficiency by 4% and 18%, success probability by 16% and 13% and power efficiency by 30% and 28% at 700 MHz and 3500 MHz frequencies respectively over ML at signal-to-interference-plus-noise ratio (SINR) = 30 dB. The analysis of power consumption with respect to threshold SINR reveals that modified GA approach reduces power usage by 18% and 14% over ML at 700 and 3500 MHz respec tively, while maintaining acceptable SINR levels, especially at higher thresholds.