Communications service providers (CSPs) are under increasing pressure to reduce their energy consumption, and this issue has been rising steadily to the top of their corporate agendas. In this report, Caroline Gabriel, Research Director of Networks at Analysys Mason explores how AI is critical to operators’ bid to reduce energy consumption in the 5G era.
There are many reasons for this:
• The current spike in fuel costs will drive up energy costs and, if action is not taken, this could put opex efficiency programmes at risk
• Consumers are increasingly concerned about the environment and interested in the green credentials of the companies they purchase from
• Governments around the world are putting sustainability at the heart of their economic plans. Many CSPs have pledged to support large-scale sustainability initiatives such as the European Commission’s 2019 Green Deal, which set a goal of becoming carbon neutral by 2050 across the bloc.
• National and supra-national goals are filtering through to specific regulatory policies in the telecoms market, such as India’s mandate for renewable energy sources to be used in rural cell towers
These combined pressures are driving a rising number of CSPs to put environmental, social and governance (ESG) issues at the very top of their long term strategic agendas. As José-María Álvarez-Pallete López (CEO of Telefónica) put it in November 2019: “The central axis of our strategy requires innovative, intelligent and sustainable technology solutions that generate a positive impact on the environment and help manage the digital transition.”
Energy efficiency is critical to CSPs’ sustainability and cost control
Reducing the energy consumption of their networks is critical to CSPs’ broad efficiency
and sustainability goals, and they increasingly recognise that this issue is now a commercial
imperative. Vague targets and lip service are certainly no longer enough to satisfy the increasingly well-informed consumer, let alone the regulator, or the operator’s own CFO.
Analysys Mason conducts an annual survey of senior executives within over 80 tier one CSPs, and the results reveal that energy efficiency has, in every year from 2018 to 2022, become a more important strategic priority, from the viewpoints both of sustainability and cost control. In 2018, energy efficiency was, on average, eighth of the CSPs’ 10 top strategic priorities for the next decade; in 2022, the figure had risen to fourth.
Within these long term roadmaps, concrete actions need to be decided now, to support short and medium term improvements. CSPs are evaluating many different ways to reduce energy consumption, from introducing liquid cooled base stations to replace air conditioning, to retiring legacy copper or wireless networks.
But perhaps the greatest opportunity to achieve long term, sustainable efficiencies lies in the intelligent management of the network itself, supported by emerging AI/ML technologies.
The mobile network has the greatest potential to deliver energy efficiencies
The telecoms industry is relatively energy efficient compared to some sectors. About 1.5% – 2.0% of all electricity usage is related to telecoms networks (a figure that is fairly stable), and a further 1.5%-2.0% is related to broader ICT. (This excludes the energy associated with network construction.) However, rising use of broadband data, and the need to increase network capacity and density to support that, threatens to increase energy consumption and cost. On average, energy accounts for about about 6.2% of total telecoms opex.
Only about 5% of CSP energy usage is not network-related (power for corporate offices, stores and field-force fleets, for instance). So the networks themselves must be the main focus of innovative energy efficiency initiatives.
[Source: Analysys Mason, 2022]
Furthermore, the focus must be firmly on the mobile network. For fixed-line operators, migration to fibre introduces a network that has significantly lower energy consumption than copper. According to Telefónica, fibre-to-thehome (FTTH) has the potential to reduce the energy consumption of fixed access network equipment by around 85%.
By contrast, the radio access network (RAN) can account for around 70% of a mobile operator’s energy consumption. Vodafone stated in 2022 that the proportion of its energy usage and costs that derives from the RAN increased from 66% to 73% between FY2019 and FY2021, an absolute rise of 16%, whereas other areas’ energy use fell by 15%. Figure 1 shows Analysys Mason’s estimate of the breakdown of energy consumption in a developed telecoms market, excluding legacy copper and 2G/3G technologies.
For the telecoms industry, then, new approaches to energy management in the mobile access networks will deliver the biggest savings, especially if CSPs focus their efforts on the aspects of the RAN that are currently least efficient
Within the RAN, according to Nokia, energy consumption is typically split as follows:
• Forwarding bits 15% (the only energy that is supporting revenue generation)
• Base station operations – broadcasts and idle resources 21%
• Fans and cooling systems 55%
• Heat, light and UPS 9%
With data transfer accounting for only 15% of energy usage, this means 85% is expended just on keeping base stations in operation.
In the run-up to the introduction of 5G in Europe, much was made of the improved energy efficiency of the 5G air interface, compared to that of 4G. 3GPP’s 5G specification, in line with ITU targets, called for a 90% reduction in energy consumption compared to 4G, on a like-for-like basis (same type and number of base stations with same traffic and signalling load).
5G NR was designed with greatly improved support for energy savings during low-to-medium traffic, and also improved the feature called microsleep TX, which puts radio units into sleep mode whenever there is a gap in transmission.
However, despite these inherent improvements in 5G NR, studies suggest that the 90% reduction has not been achieved in real world deployments.
Once the complete network is taken into account, rather than individual base stations, the impact of 5G on energy consumption and cost is seen very differently. Energy costs are increased by the use of relatively power-hungry components in some 5G equipment, such as massive MIMO antennas. Far more important is the increased number of network elements, The rise in total energy consumption and cost mainly relates to the need for 5G networks to support very high levels of traffic in heavily loaded areas, and to support the extensive coverage required for some 5G use cases such as ubiquitous vehicular connectivity.
• Base station densification (more antenna elements)
• Network densification (more cells)
• Larger numbers of spectrum bands, including high frequency bands
These all influence one another, and each one can contribute to rising energy consumption and cost across the total network. All of them will be accelerated where CSPs introduce some of the high-bandwidth services envisaged once 5G standalone is implemented.
CSPs have considerable scope to reduce energy consumption
Despite the energy challenges of implementing dense 5G, there are many way that CSPs can reduce their consumption, especially in the two main areas of energy usage, the base stations (36% of total) and the cooling systems (55%).
The key is that energy consumption depends not just on the architecture but the usage patterns, and equipment or whole sites may be idle much of the time during periods of low traffic. Many of the solutions to increase energy efficiency and reduce cost relate to intelligent management of base station power up/power down according to traffic patterns.
This is the most important way to reduce power consumption in the active base station. In Europe, for instance, 30% of base stations typically carry 80% of the traffic. This means that real energy consumption must be measured in two ways:
• Efficient data transmission in a loaded case, which relates to average spectral efficiency
• Minimising energy consumption where there is no data, which means a high sleep ratio
A study by Nokia and Telefónica2, published in December 2020, concluded that 5G networks could indeed be up to 90% more energy efficient than legacy 4G networks, but this was after Telefónica had introduced power efficiency measures such as automated power-down for idle base stations.
Automated and intelligent management and prediction of traffic patterns, and therefore of power up/down, is greatly enhanced by the use of AI/ML. Along with new methods of cooling, Analysys Mason believes that intelligent power management is the most important approach to mobile network energy efficiency, especially in 5G. That means this is also the biggest contributor to overall telecoms industry energy efficiency, since mobile networks are the biggest consumers of power and the only networks where power consumption is rising.
AI can enhance energy efficiency measures across all network domains
Intelligent power up/power down management is estimated to reduce RAN energy consumption by as much as 10% on its own, according to Analysys Mason calculations3. When combined with other AI-powered approaches to intelligent network and power management, it can reduce energy costs and carbon footprint by 30%. CSPs and towercos are starting to evaluate
AI-driven technologies for automating wakeup/sleep (for example American Tower is developing such a system itself to improve its tower energy costs and potentially offer as a value-added service). A survey conducted by GSMA and Nokia4 found that 78% of CSPs will rely on AI-based solutions to help reduce energy consumption.
AI-based energy management automation is already proving itself as a fast track to shrinking the carbon footprint of telecoms networks, especially mobile ones. The solution can predict low traffic periods and shut down resources such as frequency carriers or even whole sites dynamically. The main impact comes from automatic wake-up and sleep including base station shutdown on the basis of symbols, channels or carriers.
Nokia calculates that AI-based solutions achieve two to five times higher savings than non-AI systems that perform temporary shutdowns based on fixed schedules. Critically, AI also enables a dynamic and subtle approach to power up/down, that does not have any negative impact on performance or end customer experience, unlike some early power management tools, which could switch off service altogether for some users because thresholds were too static.
With AI, the algorithms take the specific configurations and conditions of every site into account, assuming that every site is different rather than making broad generalisations to underpin power rules.
AI can also be important to enable network density to generate efficiencies rather than boost consumption. This involves flexible, predictive cooperation between 4G and 5G radios, between different spectrum bands, and between the macro and small cell layers, to achieve the optimal balance between performance and power efficiency in any given traffic load. For instance, adding a low power 5G NR small cell to an LTE macrocell can enable traffic to be offloaded from the macro node in a quiet time, so that the macrocell can stay asleep for a long period of time.
CSPs also need to apply AI beyond the active network, since about half of energy consumption is related to auxiliary components such as fans, cooling systems, lighting and other power supplies. AI-powered energy management must cover both active and passive equipment.
Machine learning with predictive analytics can track energy usage trends with enormous granularity, to detect performance anomalies not just in the active radios, but in traditionally ‘invisible’ passive equipment, which might be draining power, perhaps because a unit is old, misconfigured or malfunctioning.
AI software solutions such as Nokia AVA can achieve savings rapidly and cost-effectively
Whatever their impact on energy efficiencies, adoption will still be limited if the technology involves significant additional cost, commercial risk or disruption to skills and processes.
Some early AI-based systems were highly complex to deploy and run and were seen as the preserve of only a few CSPs with significant internal skills and resources.
However, the new breed of AI-powered network and power management systems is very different. Software-only offerings that cover every aspect of the sites, such as Nokia AVA, can be set up within a few weeks on existing hardware, including the public cloud, with no need for associated network modernisation, architecture redesign or specialised skills. And because the solution is cloud-based, it has no dependency on a particular base station type, and can be used in all kinds of equipment, including small cells, and in multivendor networks.
Despite the simple deployment process, AVA can be scaled up immediately to manage active and passive equipment in thousands of sites and cells. And risk is further reduced by Nokia’s outcome-based software-as-a-service (SaaS) model, which enables CSPs to pay only for the energy efficiency outcomes they actually achieve.
All this means that CSPs can achieve a very quick and visible impact on their energy consumption and cost, with immediate savings generated after a few weeks of set-up time. Power savings ranging from 7% to 30% have been seen in real world, large-scale networks from day one of full operation of an AVA system.
The impact of implementing an AI-driven solution will be even more significant as mobile networks, particularly 5G ones, become increasingly dense, and as traffic patterns change with new use cases such as Industrial IoT applications. These will put different strains on the network alongside those of ever-faster mobile data connections, and it will be critical that the power management system can adapt dynamically and without needing to be reconfigured or upgraded manually. AI-based systems, then, future-proof the energy efficiency approach as well as delivering short term gains, and the resulting savings will help CSPs save cost immediately, while contributing significantly to their sustainability goals and longer term green agendas.
This report first featured inside VanillaPlus magazine. Subscribe here to get our free, quarterly digital magazine.
