NGMN says AI can help telcos plot a course to green networks

Understandably, AI features towards the top of the billing – it's fast becoming an industry-standard method for adding a dash of intrigue to any subject.

In the case of NGMN's Green Future Networks: A Roadmap to Energy Efficient Networks, AI/machine learning (AI/ML) has been identified as a key enabler for a number of energy-saving measures that telcos can take. These include predicting energy consumption, assisting MNOs with energy-saving decisions, and identifying parts of a network that require improvement.

It's worth noting that many operators have already started using AI/ML to improve the energy efficiency of their networks. For example, BT's EE arm last month became the latest UK player to begin a nationwide deployment of AI/ML-based cell sleep technology, joining Vodafone and Three, which are also either trying it out or commercially deploying it.

While AI/ML undoubtedly has an important role to play in green networking, there is a lot more to the NGMN report. It evaluates a broad range of different methods for improving power consumption, bringing them together into a single, fairly concise (by telecoms standards) guide to operating greener networks.

There are 16 recommendations, divided into three categories – process optimisation, engineering/operational improvements, and deployment of new technologies. Each recommendation has been further classified as either a short, medium or long-term priority.

The majority of them fall within the process optimisation category, which encompasses techniques and technologies for powering down various parts of a network during periods of low demand.

Engineering/operational optimisation covers things like active RAN sharing, interworking between comms networks and local power supply, and benchmarking the energy efficiency of Open RAN radio units (RUs) against traditional RAN.

Recommendations categorised under new technologies include something called front-end adaptivity. According to the report, it means the network can optimise power consumption by selecting a frequency modulation scheme and transceiver front-end based on demand and spectral availability.

Another new technology is a resource-aware machine learning framework, designed to reduce energy consumption when training large language models (LLMs).

"Reducing energy consumption while maintaining service performance is a key ongoing challenge for the industry," said Arash Ashouriha, chairman of the NGMN Alliance board and SVP of group technology at Deutsche Telekom. "NGMN's Green Future Networks Programme continues to provide the industry with valuable actionable guidance by pooling the very best of industry knowledge and shaping clear recommendations."

The full report, complete with the 16 recommendations and estimates as to how much power they could save, is available here (PDF).