PARTNER FEATURE: In the 21st century, with the development and application of emerging technologies such as artificial intelligence and big data, human productivity in science and technology has reached a brand-new stage. We have never stopped thinking about how to achieve a better balance between economic development and energy conservation and emission reduction. In 2015, the United Nations released SDG17, including Climate Action, which calls for all nations to act to protect the planet while promoting economic prosperity.

Today, let’s start with Green 5G. Let’s talk about how to balance lossless user experience and automatic energy saving, and how to implement energy conservation and emission reduction in everyone’s network life.

Green 5G: A Common Pursuit for Communicators

5G brings ultimate experience of large traffic and low latency, as well as predictable traffic growth by dozens of times. Although the energy consumption per bit of 5G networks is only one tenth of that of 4G networks, how to implement Green 5G through more effective energy-saving measures is still a challenge for the telecom industry as mobile networks enter the 5G era.

Green Limit of Mobile Networks: Bits Determine Watts

In a mobile network with wide coverage, the traffic volume is unevenly distributed in the space and time dimensions due to the difference in user habits in different areas. According to statistics, 50% of low-traffic cells on the network carry less than 20% of the traffic on the entire network. In terms of time, the wireless network is lightly loaded for more than 8 hours every day.

Let’s imagine that a base station is like an energy transmitter. On the one hand, the energy transmitter can track and even predict the mobility and network usage of people in real time. On the other hand, the energy emitted by the energy transmitter can be automatically adjusted to meet the service experience requirements of the covered people. Is it not perfect to waste no energy?

Actually, energy consumed by the base station may be adjusted by shutting down symbols, carriers, and radio frequency modules, to implement energy saving. That’s what we think of as the energy saving limit: bits determine watts.

Then, let’s think about this. If a network optimization expert is assigned to monitor the traffic fluctuation around each BTS for 7 x 24 hours and deliver the BTS energy-saving policy every few minutes, will it be crazy? Even careless giddy shut the wrong carrier, user complaints will follow, the collapse will follow. Of course, this scenario will not happen because such ultimate personalization and refined operation cannot be achieved manually.

Today, let’s see how the AI engine Power Turbo drives the wireless network to keep approaching the energy-saving limit of “bits drive watts” and maintain a lossless user experience.

How to achieve both lossless user experience and automatic energy saving?

AI-based Multi-RAT Coordinated Co-coverage Identification Enables More Base Stations to Be Included in the Energy-Saving Region

Specifically, based on site engineering parameters and a large number of terminal MRs on the network, intra- and inter-eNodeB co-coverage relationships between frequency bands are established. Inter-RAT coverage overlap relationships are identified, the proportion of co-coverage cells is increased, and the co-coverage cell identification rate is improved by 20%.

AI-based traffic forecast automatically sets the energy saving time for each site, so that the energy consumption can accurately match the traffic.

The statistics show that the network traffic volume has obvious tidal effect. For example, the traffic volume of a site in the center of an urban area varies with the commuting time of people. For sites located in residential areas, traffic peaks at night.

After the AI technology is introduced, the mAOS automatically collects historical traffic data and KPIs on the network and generates a 24-hour traffic model for each cell. In this way, the mAOS can accurately forecast the radio resource usage of each cell and control the start and wakeup time of each energy-saving cell.

AI-based personalized shutdown threshold setting, achieving in-depth energy saving in lossless experience

Operators have specific requirements for radio network KPIs. Therefore, the actual scenario information and energy-saving policies must be considered to achieve a balance between energy-saving effects and network performance. Therefore, by learning historical traffic and KPIs, the mAOS constructs an association model between energy saving and performance, continuously optimizes models through machine learning and deep learning, and dynamically adjusts energy saving parameters. In this way, the mAOS selects the optimal energy saving policy without compromising KPIs, striking a balance between performance and energy saving, and fully exploring network energy saving potentials.

Automatic intelligent energy saving in the 5G era is the basis of Green 5G. This includes data collection and learning, energy-saving region identification, energy-saving period identification, and automatic policy generation and configuration. The entire process is automated, which reduces labor costs, maximizes network energy saving efficiency, and achieves the optimal energy saving effect.

Currently, Huawei’s PowerStar energy-saving solution has been widely used at multiple sites in and outside China. For example, an operator in China uses the intelligent network energy-saving solution to reduce energy consumption by about 10%, that is, reduce 1.74–3 kWh power consumption per site. Assuming that 6.5 million 5G sites will be deployed globally by 2025, Huawei’s energy saving solution can theoretically reduce carbon emissions by 43 million tons of CO2 and save 55 billion kWh of electricity per year, equivalent to planting 380 million more trees.

AI-powered 5G networks help you and I join the energy conservation and emission reduction campaign, better protecting the planet that we depend on in our daily network life. Let us work together to see sustainable development through to the end.