ZTE Takes AI Across the Network to Drive Efficiency

Across the world, 5G networks offer increased capacity and higher speeds for consumers. More than 737 million subscribers now enjoy the next-generation mobile service in 72 countries, data from GSMA Intelligence showed. Meanwhile, monthly data usage has jumped nearly threefold compared with LTE networks.
While the new networks are significantly more efficient, delivering more bits per megawatt of energy, overall power consumption is soaring as data usage grows exponentially. Industry insiders forecast 5G energy consumption could triple by 2025.
Peter Jarich, head of GSMA Intelligence, kicked off a recent Mobile World Live webinar by saying building a green network starts by looking at your data assets to understand how and where you are using energy, and where efficiencies can be made.
"Start doing something now," he suggested, by beginning transformation planning for renewable sources and resource optimization.
In fact, AI-driven network management is increasingly part of every vendor's offering.
GSMA Intelligence research found 57 percent of operators surveyed expect network energy costs to increase by up to 20 per cent in the next three years as demand for data continues to climb, he said. "It's not surprising then that 66 per cent of respondents view energy efficiency strategies as very important," noted Jarich.
Jarich also recommended thinking beyond equipment to the entire supply chain.

Driving Efficiency

Ding Yi, senior director of analysis and strategy at ZTE, noted climate change has forced the industry to rethink how it builds and operates networks, with the mobile sector emerging as a leader in the move to net-zero emissions.
He argued that in addition to supporting more sustainable development of communications networks as data demand soars, more energy-efficient equipment can also significantly lower Opex for operators.
Research by GSMA Intelligence estimated the RAN accounts for 73 per cent of an operator's energy consumption and the core network 13 per cent.
Yi explained that improving efficiency is mostly about shutting things down in a smart way by applying AI to a cloud platform. For example, at night most RAN equipment continues to consume energy even when not active.
ZTE incorporated shut-down mechanisms and a hibernation mode to AAUs, reducing consumption at idle times to just 5 watts. Normal power consumption with low traffic is typically 200 watts, Yi said. "This is the foundation of our toolkit."
As the company investigated where in the network to use AI algorithms, its research teams realised it is was clear they should be deployed in as many places as possible.

More Than Hardware

ZTE's latest green network solution is focused on the underlying hardware in the base station, as AAUs and RRUs account for more than 80 per cent of power consumption at the base station.
The company was the first in the industry to apply AI algorithms on base stations, he claimed. "The cloud platform has some limitations. Putting AI on the BBU creates real-time responses and makes it more nimble."
Yi acknowledged AI-driven networks are not new, but deploying AI directly to the base station is a recent innovation. "Cloud platform AI is the norm in the telecoms industry at present, but power saving using local AI at the base station is truly innovative."
Based on big data analysis, the platform automatically identifies network energy saving opportunities, predicts network traffic trends and identifies co-coverage cells. The system also automatically generates energy saving policies and implements network-level intelligent energy reduction management and site energy-saving coordinated scheduling.
ZTE takes a three-layer approach to improving energy efficiency: starting with the hardware, using in-house developed chips and powering its algorithms with advanced software.
With improvements in computing power used in base stations, more intelligent strategies can be introduced, allowing cross-domain information affecting consumption to be automatically generated at the base station to improve optimization.
Nearly 20 operators, including China's major three mobile players, have deployed its PowerPilot Pro on some 900,000 base stations running on more than 30 networks globally. 
Earlier in the year, Chengdu Telecom partnered with ZTE to use the vendor's AI intelligent solution to reduce energy consumption, potentially cutting consumption as much as 34 per cent on its 5G network.

Core Gains

ZTE technical director of cloud core networks Yang Rui insisted focusing on the core network can make a big difference in an operator's green transformation efforts.
"Although the energy consumption of core networks only accounts for about 13 per cent of telecoms networks, there are still many parts that need to be optimized," she stated.
ZTE's goal is to continue improving system performance—using the minimum resources for the maximum capacity, deploying full virtualization to lower the watts used for each bit of tariff, she said. Being service aware is also vital, requiring the system to apply a 'green brain' for energy saving policies based on service perception.
To reach these objectives requires not only open systems in order to collaborate with partners but a cloud native platform.
Yang explained a cloud platform, more than ever, plays a significant role in overall network performance. "It provides flexible adoption of different functions, like data forwarding and control signalling processing. For different kinds of applications, we can assign different resources and adjust the network setup."
The platform supports a dynamic power saving mechanism to automatically shut down some resources when there's no payload and can run defragmentation to release extra resources that have been consumed.
A heterogeneous cloud infrastructure provides the most efficient resource for applications, she added. In addition, unified management and orchestration of different processors also depends on the perception of service and global resource orchestration.
These intelligent energy management systems also can be applied to network slices, which means they can be used across different types of data centres.

Smaller Is Better

Liu Xiao Guang, chief engineer of Digital Energy Planning at ZTE, explained that simplifying telecoms sites can dramatically reduce power consumption, for example, by fitting the air conditioner, power supply, batteries and transmission into one cabinet. ZTE has done the same with its pole sites.
For data centres, cooling and power distribution consume the vast majority of the energy. Liu noted new technologies such as indirect evaporative air handling units (AHUs) and liquid cooling can reduce power usage effectiveness (PUE).
“If we want to move closer to our carbon neutrality targets, more renewable energy, coupled with intelligent energy management, needs to be used to power hungry 5G networks. And energy storage systems will make renewable energy more reliable and useful,” Liu stated.