Article Details

Abstract

To address the energy efficiency bottleneck problem in 5G base stations under low-load scenarios where hardware resource utilization is less than 30%, this study aims to reduce static power consumption and improve green energy-saving performance through dynamic regulation. An optimization scheme based on Dynamic Voltage and Frequency Scaling (DVFS) is proposed, and a three-tier “Perception-Decision-Execution” architecture is built. The perception layer employs a multi-dimensional hardware monitoring system with the 10  kHz sampling rate, the decision layer integrates Deep Reinforcement Learning (DRL) and the LSTM-based load prediction algorithm for real-time Vdd-f configuration optimization. The execution layer incorporates a programmable hybrid Power Management Unit (PMU). Experimental validation on a semi-physical simulation platform shows power reductions of 25.8%, 27.4%, and 31.4% under idle, low-load, and tidal-load scenarios, respectively, with a 59.7% improvement in overall energy efficiency and QoS compliance maintained between 97.8% and 100%. This research provides a scalable and efficient technical framework for dynamic energy efficiency management in 5G base stations.

Keywords

Dynamic Voltage Frequency Adjustment (DVFS); 5G base station; low load power consumption; energy efficiency optimization

References

1. [1] Liu Jingling, Huang Jiawei, Jiang Wanchun, etc A Review of Load Balancing Methods in Data Centers [J]. Journal of Software, 2021, 32 (2): 300-326
2. [2] Zhang Jing Virtual Machine Merge and Deployment for Energy and Heat Perception in Cloud Data Centers [J]. Computer Applications and Software, 2021, 38 (1): 63-69104
3. [3] VANEET G, BALKRISHAN J. Efficient Virtual Machine Migration Method for SLA Conservation in Cloud Computing [J]. Journal of Central South University, 2021, 28 (3): 760-770
4. [4] Tang Lun, Helanqin, Tan Qi, etc Virtual Network Function Transfer Optimization Algorithm Based on Deep Deterministic Policy Gradient [J]. Journal of Electronics and Information Technology, 2021, 43 (2): 404-411
5. [5] Yang Lina, Zhao Peng, Wang Peizhe Research on Data Center Energy Consumption Prediction Model Based on GRU Neural Network [J]. Electric Power Information and Communication Technology, 2021, 19 (3): 10-18
6. [6] Zhao Yuyu, Cheng Guang, Liu Xuhui, etc Overview of Next Generation Network Processor and Applications [J]. Journal of Software, 2021, 32 (2): 445-474
7. [7] Queqing River A Low Cost and High Performance RISC-V Processor Design [J]. Microcontrollers and Embedded System Applications, 2021, 21 (1): 14-18
8. [8] Duan Weihu, Lu Haitao SOC architecture analysis based on DVFS technology [J]. China Integrated Circuit, 2020, 29 (12): 73-77
9. [9] Wang Yichen, Wang Lunyao, Xia Yinshui, etc FPRM logic power optimization based on approximate computing techniques [J]. Journal of Ningbo University: Science and Engineering Edition, 2020, 33 (1): 45-50
10. [10] Zhao Shan, Yang Qiusong, Li Mingshu Heterogeneous aware scheduling technology under asymmetric multi-core processors [J]. Journal of Software, 2019, 30 (4): 1164-1190

About the authors

Li Wen (1982-), male, Han nationality, from Loudi, Hunan, engineer, master's degree, research direction: enterprise management, project management, communication engineering.