Improving energy efficiency in networks by AI

The global energy consumption for cellular networks is a hot topic. The carbon footprint and the costs associated with energy consumption are one of the biggest challenges facing our industry. As long as the energy mix used by the networks is not climate-neutral and partly comes from fossil fuels, mobile communications contribute to the emission of greenhouse gases into the atmosphere. This means that reduced energy consumption through more intelligent use of cellular networks can have a smaller impact on greenhouse gas emissions.

If energy consumption continues to rise, energy prices are likely to rise too. In any case, there are higher costs for network operation and the combination increases the total cost of ownership. Combating energy consumption requires a concerted effort from the telecommunications industry. It has to be viewed from every angle and must leave no stone unturned.

An inconvenient truth: for RAN

The first step in solving a problem is to accept that there is one. This is why our network analytics function supports the observability of energy consumption over the generations. The main villain of this story is site consumption. Around 85 percent of network consumption is accounted for by Radio Access Network (RAN) locations, the remaining 15 by data centers. This could increase in the case of cloud RAN.

The upcoming Intelligent Automation Platform from Ericsson will support rApps as well as the future Energy Saving Manager. The Energy Saving Manager can decide centrally which energy saving functions are activated and through which configuration.

Part of the Ericsson philosophy for intelligent RAN automation is to choose centralized or decentralized control for each specific use case in order to achieve maximum impact. For improved energy efficiency, the central control enables holistic decision-making based on the analysis of data from several locations. This analysis can then be used to create the best local setup to maximize KPIs and performance while reducing overall energy consumption.

Improved data collection and analysis goes one step further, enabling service providers to make active decisions about prioritization. This also includes considering whether (and how much) negative KPI effects can be allowed in order to further improve energy savings. This can vary depending on where in the network energy savings and KPI targets can be achieved and adjusted.

Think globally, act locally: on the knot

In order to save energy at the node level, the energy measurement function must support the energy analysis function. MIMO Sleep is an efficient feature to keep the user experience while minimizing waste when less capacity is enough. The problem was that it used to require manual configuration, which is both time consuming and less efficient. To address this problem, we introduced AI-powered MIMO Sleep, which automates parameter setting to reduce manual work and improve the performance of functions (both for KPIs and energy savings) at the node. Further information can be found in the results of the PoC.

Features like the AI-powered MIMO sleep mode allow us to make the most of the current paradigm of resource use. The next leap in node energy savings will come through a paradigm shift. Nowadays resources are by definition “always on” and are switched off or put into energy-saving mode when the traffic situation permits. Going forward, we envision a transition to an “always available” paradigm where resources rest until they are needed. With intelligent predictions, they can be activated if necessary. With AI-powered capabilities, we can make accurate traffic forecasts for further savings while allowing users the same great performance.

The day after tomorrow: an outlook

Looking ahead, we see that reinforcement learning (RL) approaches will further improve energy savings and network performance. RL is particularly useful in the type of dynamic, complex, and high-demand environments that make up mobile networks.

There are several ways in which RL can be applied to networks in general and energy saving in particular. An example are the two successful attempts Ericsson has completed to apply reinforcement learning to Remote Electric Tilt of Antennas (RET). At first glance, it doesn’t seem that complicated, but every time you tilt an antenna it changes the shape of the cell that the antenna is in. This in turn influences the user experience of the people served by this cell and the cells around it and is further tilted by surrounding antennas and acts cascading in the network. This makes it all the more impressive that Ericsson and the partner service provider in a single live network when optimizing for reduced ERP caused a 20 percent reduction in DL transmit power without compromising performance.

RL also offers more options for extensive and complex orchestration. For example, energy savings could be built into traffic control and used to direct traffic to the most energy efficient resources on the network. This would allow other resources to sleep deeply while the traffic control scheme is active.

If we take a step back and look at the network and its entire life cycle, it is clear that we should not only optimize for today and what is already in use, but should also seriously consider what is to be used in the future. Thinking about what and where to deploy is especially important now that much of the world is in the midst of the 5G rollout.

It’s simple: smarter, more accurate deployment reduces the hardware required and the network’s environmental footprint. The cognitive software suite has several features that support this. Including capacity planning for traffic forecasting, site selection to determine the best location for deployment, and RF design to optimize the models used for network design. The overall effect is an optimized network layout that reduces the pressure on the wallet and the environment.

Earth hour, every hour

At Ericsson, we know that business can only thrive in a sustainable environment. But in contrast to the earth hour, cellular networks must of course not be switched off, so we all have to make sure that we optimize every hour. We take this challenge seriously and offer solutions for all cellular networks and their life cycles. We believe this will benefit us and future generations.

Contact us if you would like to find out more about how we can help you reduce the carbon footprint of your cellular network and become more energy efficient!

Related content

Intelligent RAN automation

AI: Enhancing the Customer Experience in a Complex 5G World

An intelligent platform: The use of O-RAN’s SMO as an enabler for openness and innovation in the RAN area

Secure energy-efficient networks with artificial intelligence

Can AI reduce energy costs in the network?

Ericsson introduces AI-powered energy infrastructure operation

Energy efficiency

How to break the energy curve

Leave a Comment

Your email address will not be published. Required fields are marked *