Trending

Energy efficiency: AI for mobile networks


“Over a large perimeter, these algorithms are capable of identifying sites that are atypical compared to a reference site.”


On the same subject

The new mechanisms introduced by 5G improve the energy efficiency of networks and optimise their carbon footprint. The combination of Big Data and AI technologies, for example, gives us a glimpse of significant prospects for progress for less polluting mobile networks.

Each generation of mobile telephony networks coincides with a leap forward in energy efficiency per bit transported, the gains of which are estimated at a factor of 10 during the transition from one generation to the other. 5G is no exception to this rule and, from the infrastructure design stage, introduces new levers for technical and operational optimisation. Instead of waiting for these changes, carriers are developing solutions to minimise the carbon footprint of current 3G/4G networks.

These efficiency levers are already paying off

In this regard, Orange is working in the long term to reduce consumption and CO2 emissions linked to its infrastructure, products and services. In particular, its Engage 2025 strategic plan aims to reduce the Group’s CO2 emissions by 30% by 2025 (compared to 2015), moving towards a long-term “Net zero carbon” objective by 2040. The Green ITN plan, which has been deployed for several years, is a major building block of this ambition and focuses on Orange’s main source of energy consumption: networks and information systems. “Significant efforts have been made within the framework of the Green ITN 2020 plan,” says Quentin Fousson, Green RAN & Benchmark project manager. The identification, design and deployment of some fifty levers for optimisation have, over the past ten years, prevented 2.7 million tonnes of CO2 emissions and reduced the energy costs for our networks by one billion euros. In recent years, Orange has managed to stabilise the amount of CO2 emissions linked to the operation of its networks. The Green ITN dynamic plan is now projected for 2025 and new areas of work such as the energy efficiency of 5G and the eco-design of data centres are already under way.

Measure, analyse, optimise

In recent years, Orange has deployed meters on a large scale in the countries where it operates to collect a large amount of data on the consumption of its network infrastructure. The dynamic visualisation and analysis of this data, with Big Data technologies, are very valuable vehicles for rationalisation in the monitoring and operation of current networks. This data can then be used for the targeted implementation of energy efficiency solutions, such as the application of standby mode when and where possible, to limit consumption when traffic is slowed down.

Using AI to systematise

This work continues with the research and development of new improvement methods accelerated by artificial intelligence techniques. “We are using AI to take our optimisation even further,” said R&D engineer Yuanyuan Huang. Algorithms have been developed to review the masses of data available to us, and compare the performance of radio sites on traffic delivered to the customer and their energy consumption. Over a large perimeter, these algorithms are capable of identifying sites that are atypical compared to a reference site. In the event of too great a discrepancy — if one site has a higher consumption than another site with an equivalent volume and quality of past traffic, for example — energy efficiency recommendations may be issued for the sites concerned: reduction of the power emitted by the antenna or placing transmitters on standby during periods without traffic.” 

Other applications being studied

This is just one of several use cases studied in the Group’s different countries today. AI also makes it possible to detect oversized sites or to identify time windows to which standby mode can be applied. A dozen use cases were explored, with average savings in network consumption and CO2 emissions ranging from 2% to 10% depending on the use case and the network architecture in the subsidiaries.


“Over a large perimeter, these algorithms are capable of identifying sites that are atypical compared to a reference site.”


On the same subject

Keyword :

, , , ,