Driving Energy Efficiency in 5G Networks Using AI
5G infrastructures will consume significantly more power than the 4G systems they replace. This is in part due to the inexorable growth of user data supported by the much greater capacity available through 5G technologies, but also due to the larger number of base stations deployed to support this capacity. As a result, over 90% of leading mobile network operators have expressed concerns about the rise in energy costs they have to sustain .
The Radio Access Network (RAN) part of the 5G infrastructure consumes a significant fraction of this power to generate the radio signals that allow wireless communications with the end user devices. This has traditionally been the case in previous generations of cellular networks , but will be exacerbated in the case of 5G systems: for instance, according to recent assessments, energy consumption in the edge of softwarized mobile networks may increase as much as 25% solely due to the impact of active cooling , , which sums up to the much denser BS and/or Remote Unit (RU) deployment that high-performance but low-range technologies like mmWave require.
Operators are then called to make energy savings a higher priority than in past, in order to both guarantee the sustainability of mobile communication infrastructures and the compliance with green recovery programmes. Indeed, major operators believe that a cultural shift that moves sustainability and energy efficiency at the core of corporate strategy will be one of the top three solutions to drive substantial Operating Expenses (OPEX) in 5G technologies . Achieving such a change requires a more holistic approach to energy-prudent operations that goes beyond current approaches, which often leave power management solutions to individual operating units.
In this context, the predictive automation of user behaviour can significantly reduce the power consumption of 5G RAN both through optimization of the compute resources needed to support data transmission and through intelligent management of the radios, channels and amplifiers at the heart of 5G radio towers. Such an intelligent management of the RAN throttles network capacity based on the fluctuations in the demand and is key in ensuring an optimal balance between the two conflicting goals of energy saving and network quality. According to GSMA, two thirds of operators expect Artificial Intelligence (AI) to drive savings ranging from 10% to more than 30% in the next two years .
“Increasing energy efficiency is one of our top company goals, and we focus a lot on this topic. The problem with intelligent management of radio resources (i.e. reducing network capacities during hours of low demand) is maintaining an optimal balance between saving energy and delivering network quality to our customers. We expect the main benefit of Net AI’s ML/AI technology to be automatically providing the optimal balance between these two contradicting goals in real time”.
— Thomas Hodi, AI Product Manager | Senior RAN Expert, A1 Telekom Austria Group
Using its innovative AI driven data analytics, Net AI can predict network usage on an application-by-application basis — allowing operators to attain energy savings in two ways. First, Net AI models can produce optimal periodic (e.g., hourly or daily) configurations of the RAN that can be simply enacted by the operator at the corresponding time, and that potentially abide by the needs of the local user base (e.g., in terms of expected usage of specific classes of mobile services). Second, Net AI develops advanced solutions to forecast demand in real time and support the live optimization of the network performance, via a fully automated configuration of the RAN. These approaches can help reduce the carbon footprint of an ever expanding 5G network while delivering expanded capacity and capability, and contribute to meeting the ambitious targets set by major operators like Verizon, Telefónica or Vodafone, which aim at 50%-70% carbon intensity reduction by 2025–2030.
 Mobile Europe, ‘More that 90% of operators concerned about 5G energy costs,’ https://www.mobileeurope.co.uk/press-wire/more-than-90-of-operators-concerned-about-rising-energy-costs-for-5g-and-edge
 J. Wu, Y. Zhang, M. Zukerman, and E. K. Yung, ‘Energy-efficient base- stations sleep- mode techniques in green cellular networks: A survey,’ IEEE Communications Surveys & Tutorials, vol. 17, no. 2, 2015
 M. Masoudi et al., ‘Green Mobile Networks for 5G and Beyond,’ IEEE Access, 7, 2019
 E. Ahvar, A.-C. Orgerie, and A. Lébre, ‘Estimating Energy Consumption of Cloud, Fog and Edge Computing Infrastructures,’ IEEE Trans. Sustain. Comput., 2019
 GSMA, ‘The Mobile Economy,’ 2022.