2021-11-17-ngmn-绿色未来网络—网络能源效率白皮书_56页_1mb

Network Energy Efficiency White Paper Summary (NGMN Alliance)

This document, from the NGMN Alliance, focuses on improving the energy efficiency (EE) and energy performance (EP) of mobile networks, particularly 5G, given increasing traffic demands and energy costs.

Key Objectives & Goals

• Reduce Energy Consumption: Lower OPEX and CO2 emissions despite rising traffic. Use advanced technologies and network optimization.
• Enhance EE: Achieve significant improvements (factor of 2000 projected by NGMN by 2030+ for traffic growth factors).
• Role of 5G: Explain how 5G, despite higher deployment energy, offers better spectral efficiency (due to Massive MIMO, lean design) than 4G.
• Lifecycle Approach: EE considerations span equipment design, air interface protocols (3GPP NR power saving), site infrastructure, and system-level management (including sleep modes).

Major Sections & Findings

1. Energy Efficiency Techniques:

   • Base Stations (Equipment Level): Largest consumer (~57% or ~50%). Focus on power amplifier efficiency (targeting >50% in 5G), processing power reduction under low load, virtualization of RAN (using COTS hardware/accelerators), sleep modes, and efficient hardware design.
     • Sleep Modes: Granularity varies (# symbols/# time). 1st level (Symbol/μDTX ~10% savings) is mature; deeper levels require careful trade-offs with user experience (UX).
   • Site Level: Cooling (50% potential savings via Free/Liquid cooling, heat reuse); Efficient AC/DC power conversion, battery improvements, higher voltage DC distribution (400VDC).
   • Network Level: Traffic-aware deployment, virtualized functions (VNFs), HetNet management, optimizing use of diverse spectrum bands, careful multi-RAT network design to minimize idle carriers.
   • Standards: Leverage 3GPP NR features like Massive MIMO, lean carriers, improved sleep modes.

2. Impact of AI/ML:

   • Enabling Automation: AI assists in traffic prediction, shutdown coordination, and optimization across network scales without impacting QoE.
   • Energy Savings: Case studies (e.g., Telenor) show significant OPEX reduction (e.g., 3.6% network power savings) and performance improvements.
   • Energy Cost: AI's own computational needs must be considered (e.g., ~0.03 Wh per eNB/gNB per year for inference), though training is less relevant.

3. Other Key Points:

   • Device Role: UE receiver sensitivity affects overall network EE. Better device sensitivity leads to less retransmission and network power.
   • Phasing Out Legacy Tech: Sunset of 2G/3G requires strategic decisions balancing spectrum reallocation and energy/coverage impacts.
   • Server Virtualization: Reduces power usage via workload consolidation and resource optimization.

Recommendations & Conclusions

• Mature Deployment: Actively deploy established sleep/silence modes (Symbol/Channel Shutdown).
• Focus on Deeper Modes: Explore AI-driven activation/deactivation for deeper sleep modes and carrier/power shutdown, optimizing the UX/power trade-off.
• Site and System-Level Focus: Implement efficient site power/cooling (Free/Liquid + AI) and advanced network architecture (400VDC, HetNets) for EE gains.
• AI Investment: Leverage AI/ML systematically for self-optimization across all EE domains.
• Industry Collaboration: Need for joint advancement of EE technologies, better-aware rating systems rewarding efficiency alongside performance.

Annexes (Briefly)

• Executive Summary: Overview of challenges, goals, means (5G features, sleep modes, virtualization, AI, site optimization), and potentials.
• Document Introduction: Background, definition of energy related terms, and EE metrics explainers.
• References: Cite numerous relevant standards bodies (ETSI, 3GPP, ITU-T/ITU-R, NGMN, TIP), studies, and white papers.

Overall

The white paper comprehensively outlines existing and emerging technologies to tackle the energy challenge in 5G (and beyond), stressing scalable optimization, AI integration, and multi-layered strategies across equipment, site, and network levels.