Ericsson | Sustainable networks.

The Intelligent Operations guide 1

Sustainable
networks.
The Intelligent
Operations guide
2 Ericsson | Sustainable networks. The Intelligent Operations guide

Content

01. Our journey to break the energy curve ............................................................................................................4

02. The role of AI in RAN optimization ...................................................................................................................6

2.1 Accelerating AI in RAN for energy performance: The Telenor project..................................................................... 7

03. Optimizing Energy Performance ................................................................................................................... 10

04. Ericsson´s versatile energy solutions adapted to CSPs needs................................................................. 13

4.1 Optus journey to improve energy performance with automation..............................................................................15

05. How Programmability can make networks more sustainable ............................................................... 17

5.1 5G Advanced software is a starting point to programmable networks.................................................................19
5.2 The three layers to improve energy performance ........................................................................................................20
5.3 Telstra’s commitment: Enhancing user experience with sustainability................................................................23
5.4 DNB is enhancing Malaysia´s 5G network with Automated Energy Saver .........................................................25
5.5 Swisscom 5G journey commitment to performance and sustainability ...............................................................27

06. Services powered with AI to maximize energy savings........................................................................... 29

6.1 Intelligent RAN Power Saving solution with a tailored approach...........................................................................30
6.2 Umniah AI-based solution to improve energy efficiency in Jordan........................................................................31
6.3 Energy Efficiency Map...........................................................................................................................................................33
6.4 Predictive Cell Energy Management Solution for multivendor RAN......................................................................34
6.5 TDC NET deploys Predictive Cell Energy Management for energy performance .............................................36

07. Site solutions innovations beyond energy efficiency................................................................................ 38

08. Key takeaways and recommendations......................................................................................................... 42

Ericsson | Sustainable networks. The Intelligent Operations guide 3

Why operate intelligently.

How can we make mobile
networks able to save energy
and improve user experience?

Energy cost has become in the last years one of the highest This call for change brings great opportunities to CSPs. It is
operating costs for communications service providers (CSPs). possible to improve user experience while reducing energy
GSMA estimates that energy accounts for 45% of network consumption. In this report you will discover the solutions
OPEX and 15–20% of total operational spend for an average that leading CSPs are implementing in their journey to high
telco [1]. This makes reducing energy consumption a target performing programmable networks, more autonomous and
for both sustainability and network operation ambitions, and energy conscious.
in the long term, a strategic criterion for CSPs evolution plans.
We are pushing the technology and hardware boundaries for
However, many CSPs still hesitate to activate energy- a conscious energy use and user experience targets.
saving software solutions. This is due mainly to the Our vision of programmable networks offers a dynamic and
potential risk of impacting traditional network performance flexible framework that allows for real-time adjustments
indicators and the complexity of the features configuration and optimizations based on current network conditions
(based on multiple parameters and thresholds). With the and demands. By leveraging 5G advanced software [2]
traditional approach, CSPs will not be able to control energy and automation powered with AI, these networks can
consumption to achieve their targets. CSPs must evolve how intelligently manage resources, ensuring that energy
to plan, deploy, and run mobile networks. consumption is minimized without compromising
performance or user experience.

“Sustainability and energy efficiency are key for designing mobile networks today.
The programmability of the high-performing network will enable its evolution into an intent-
driven, autonomous system, capable of supporting diverse service needs. This transformation
will be powered by AI”.

Erik Ekudden
SVP Head of GF Technology & CTO, Ericsson

“Just like an organism, networks will never waste their resources and energy. When resources
are not needed, they’ll sleep to preserve energy. These networks will be intent-based saving
energy and securing seamless user experience at the same time”.

Sibel Tombaz
Head of PL Cloud & Purpose-built 5G RAN, Ericsson

1. Telco AI: State of the Market, Q2 2024

2. Explore the latest solutions of Ericsson 5G Advanced
4 Ericsson | Sustainable networks. The Intelligent Operations guide

01 Our journey to break

the energy curve

In the Ericsson report “On the road to breaking the energy curve” we provide recommendations
on how mobile networks are planned, built, and operated with precision and sustainability in
mind. Our approach consists of three steps or core pillars:

• Sustainable network evolution

• Expand and modernize
• Operate intelligently
Ericsson | Sustainable networks. The Intelligent Operations guide 5

Sustainable Operate
network evolution intelligently
Embrace a holistic Leverage AI/ML
view for network and automation
planning and to boost energy
operation savings

Breaking the
energy curve

Expand and modernize

Modernize existing network
while scaling up 5G

Total Network
Energy consumption Traditional roll-out Breaking the energy curve approach

5G
4G

3G

2G

Network roll-out over time

Figure 1: Ericsson´s approach to break the energy curve

In the report “Sustainable networks. The RAN modernization By operating intelligently, CSPs will leverage on the traffic
guide”, we have described how effective modernization of variation to save energy by adapting the capacity to the
existing networks is essential when scaling 5G to reduce the current demand to prevent the waste of resources.
mobile network total energy consumption with the latest AI and Automation technologies will also simplify network
hardware and software technology innovations. operations and boost optimizations at scale. In our approach,
we follow three principles for sustainable operation:
The evolution of our tailor-made Ericsson silicon has resulted
in enhanced energy performance enabling better sleep • Applying network insights to use energy saving
modes and increased processing capabilities. Additionally, functionality more optimized.
our efficient power amplifiers utilizing GaN technology • Connecting and getting data from all parts of the network,
and advanced filter technology have significantly reduced including passive infrastructure elements such as rectifiers,
internal energy loss. Effective heat dissipation management batteries, and HVAC systems to enable the use of energy-
from critical components has been achieved through the saving functions throughout the network.
development of innovative thermal solutions that don´t • AI/ML automation to boost energy savings while securing
require fans. Incremental energy savings are anticipated with performance.
the integration of hardware-near algorithms to fine-tune
hardware components. From a RAN software perspective, In this report “Sustainable networks. The intelligent RAN
multiple traffic-aware features have unlocked additional operations guide”, we describe our view on transforming
daily savings per radio unit while securing user experience. networks operations while evolving networks to be
programmable. With Programmable, intent-driven networks,
Ericsson is allowing operators to achieve both energy
performance and sustainable growth.
6 Ericsson | Sustainable networks. The Intelligent Operations guide

02 The role of AI in RAN

optimization

Energy performance is one of the key areas where AI is Based on Ericsson’s long experience on working with AI,
applied in mobile networks [3]. The complexity of a multi- there are three main categories of use cases where AI adds
objective scenario such as reducing energy consumption and unparalleled value and high ROI:
keeping user experience, trade-offs balancing, and using the • AI unique uses cases. There are unique corner case
traffic predictions to prevent non desired impacts, makes scenarios where AI tools will give added benefits, which
AI based solutions well suited to improve both network include prediction, anomaly detection, analysis, and so on.
performance and energy efficiency.
• High complexity use cases. One example is the link
To save energy at a radio level, operators use several adaptation feature. This functionality is designed for
techniques, the most frequent strategy is to activate radio situations where there are multiple factors to be considered
energy-saving features such as: Cell Sleep Mode, MIMO Sleep to find the best signal modulation for optimal transmission.
Mode and more. These features need to be activated and • Multi-objective use cases. These use cases involve seeking
deactivated according to the dynamic changes in traffic per an optimal solution under complex constraints and trade-
sector and per site, which entails a huge complexity. offs. For instance: handling handover choices while
maintaining certain network KPIs and maximizing energy
To understand better this complexity, let´s look at a single savings while improving user experience.
radio base station site of a modern CSP. This hub of
connectivity houses radio units that deploy distinct “sectors,”
each radiating targeted radio waves towards specific AI driven automation will be critical in the high
areas. These sectors employ unique “carriers” “ for different complexity use cases such as multifrequency network
generations of radio technologies, that also seamlessly can deployments and the multi-objective use cases
switch between technologies (4G, 5G) to deliver optimal such as improving user experience and reducing
mobile service. The same radios can also be used for earlier energy consumption. The predictive capabilities of
2G and 3G technologies. From a performance management AI based solutions can boost energy savings without
perspective, analyzing our physical site’s traffic reveals compromising user experience.
distinct patterns across different times of day (day, night,
evening). Zooming out to a weekly or monthly view, we Moving towards AI-native RAN unlocks significant
see seasonality impacting traffic, with fluctuations across advantages:
months. Therefore, we need a dynamic solution to adjust in • Scalability: AI can handle a growing number of
real time to the changes in traffic and to the types of service. complex intents efficiently, unlike rule-based systems
that can´t manage higher-levels of complexity.
Communications Service Providers are keen to reduce
energy consumption in the active and passive equipment • Optimized decision-making: AI can make real-time
decisions based on vast amounts of data, leading to
without impacting user experience and network resilience. AI
more efficient and optimal network operations.
technology will be essential to achieve these improvements in
the more complex scenarios.

3. AI journey in RAN report

Ericsson | Sustainable networks. The Intelligent Operations guide 7

2.1 Accelerating AI in RAN for energy

performance: The Telenor project
In a move to advance AI and ML technologies in the In the solution, we have developed an AI agent controlling
telecommunications industry, Telenor and Ericsson Research LTE cell sleep mode parameters of L26 in multi-band NR
have established a three-year research partnership to NSA every quarter-hour in a closed loop based on ENM
explore, develop, and test cutting-edge AI/ML solutions for observability. With reinforcement learning, Sim2Real,
enhancing RAN energy efficiency. Under this partnership, world model, and safety shield as key components of the
intelligence, the AI agent is trained using simulations and
two collaboration projects have been initiated. The first
Telenor Norway data.
project focuses on developing an AI agent to control RAN
energy-saving features in a closed loop. The second project, The solution is successfully demonstrating reduced network
as a continuation of the first one, aims to apply trustworthy energy consumption and is ready for live testing. The
AI with a focus on explainable AI and causal AI. progress achieved reflects the development and deployment
of advanced ML technologies and careful risk management,
enabled through the strategic innovation partnership
AI agents increase energy savings while keeping
between Ericsson and Telenor.
performance
All energy savings were achieved without negatively
The joint team has identified a specific use case for the AI
impacting network quality, as measured by customer
agent: a closed-loop control of parameters for LTE cell sleep
throughput. An energy consumption saving of 4% is achieved
mode, managing the dynamics between parameters and
for the L26 radio cell when the agent is handed over control
KPIs across a multilayer network, and adapting to site traffic
of cell sleep mode (compared to best practice). Time in
patterns at different times of the day.
active sleep mode increases from 4 to 7 hours on average.

RAN NR
features Capacity
(N35)
LTE
CSM LTE
AI training pipeline Capacity
(L21, L26)
…

LTE
Coverage
(L08, L18)
Simulations

Centralized closed loop

Calibration ENM PM/CM data

Data

Figure 2: AI agent implemented to configure LTE energy saving features

8 Ericsson | Sustainable networks. The Intelligent Operations guide

This use case and preliminary results on energy-throughput Creating a trustworthy AI framework with explainable AI
performance were presented in the GSMA ‘Telco AI state of
the Market’ Report. Once validated in trial, the method will be The second project with Telenor aims to develop the
expanded to control more frequencies beyond L26. foundations for a Trustworthy AI framework in the
Telecommunication industry to enhance the reliability
The figure 2 illustrates the centralized closed loop to control and transparency of AI/ML systems deployed in networks
where the AI agent controls the cell sleep mode activation of thus enabling and enhancing customer trust. It consists of
the radio cells based on real-time network performance and
Explainable, Causal and Human-Centric AI work to improve
state, adjusting parameters dynamically. Cell sleep mode is
the trustworthiness of AI-driven solutions for network
enabled in the radio cells during low activity periods, handing
automation by applying explainable and causal AI methods
over traffic to other radio cells within the coverage area.
and then validating user trust in AI.
The AI agent is trained, continuously learns from millions
of trial-and-error attempts in a network simulator with The objective is to provide explanations of the AI agent’s
historical data and real-time network information such as recommended actions. Explaining the factors that led to
performance metrics, throughput, and load. The AI agent the AI agent’s decision is expected to help users assess the
learns from its actions and the corresponding effects on correctness of the agent, thereby increasing trust in the AI
network quality, ensuring a continual improvement cycle. itself and facilitating its further adoption.

This work complements the first project by extending it

The following figure shows the performance of the different
with explainability for reinforcement learning and causality
agents evaluated and showcases the trade-offs between
methodologies.
energy consumption and throughput performance. The graph
shows the gains achieved in terms of sleep time in hours This work focuses on three main activities:
in relation to the probability keeping the user throughput.
• Explainability: Development and application of
The agents evaluated fall into three categories: Optimal for
Explainable Reinforcement Learning (XRL). With data
energy savings and performance (dark green area), optimal
evaluation, explainable RL methods will provide insights
for performance but suboptimal for energy savings (blue into the agent’s behavior such as why a particular action
area), optimal for energy savings but suboptimal for keeping is taken, the conditions that trigger certain actions and the
performance (light green area). key factors influencing decisions, and the identified corner
cases.
• Causality: Development and integration of Causal
Modelling methodologies. By observing the data from the
field, the causal mechanism identifies causal relationships
in data and generates counterfactual, adding more
insights.
• Human-Centric AI: Assessment of the drivers of
human trust in AI. All the obtained insights, obtained by
implementing explainability and causality activities, are
evaluated by an engineer.

Ultimately, with the result of this evaluation, the AI model

will be improved.

Figure 3: AI agents’ performance evaluation.

Ericsson | Sustainable networks. The Intelligent Operations guide 9

Customer

Policy Explanation
RAN NR
features Capacity
(N35)
Requirement AI agent
LTE
Policy CSM LTE
Capacity
improvment
Insights AI training pipeline (L21, L26)

…
LTE
Coverage
Interactive synthetic Field data (L08, L18)
data collection collection
Simulations Centralized closed loop

Explainable AI

Calibration
ENM PM/CM data
Data

Figure 4: Elements to create a trustworthy AI model

“We believe that co-creating innovations with Ericsson Research and Development on energy
efficiency and state-of-the-art AI positions Telenor to harness the value potential of our
operational data. This collaboration supports us in reaching our ambitious sustainability goals,
while also advancing industry standards on safe and trustworthy AI systems for network
automation.”

Ieva Martinkenaite
SVP, Head of Research & Innovation, Telenor
10 Ericsson | Sustainable networks. The Intelligent Operations guide

03 Optimizing Energy
Performance

The target for network optimization cannot be defined solely align network capacity with traffic demand. The challenge
as reducing energy consumption. CSPs are not interested in lies in transitioning from individually tuned energy-saving
purchasing equipment to switch it off; they need to provide features to an orchestrated approach that allows the network
continuous service and coverage. To achieve optimal to follow the capacity demand curve, delivering the desired
performance, CSPs aim to adjust installed capacity to real- performance without wasting energy or requiring additional
time traffic demand without affecting user experience. hardware.
Network capacity is typically deployed to handle peak
The figure below illustrates how radio resource optimization
traffic during the busiest hours, resulting in over-capacity
can synchronize capacity to the instantaneous traffic load [4].
for most of the day. Software functionality is employed to

Pre-set or manual capacity optimization AI driven power saving aligned with traffic
Traffic load
Traffic load

Network Traffic load

Capacity/power consumption

00.00 12.00 24.00 00.00 12.00 24.00

Time Time

Figure 5: Optimizing energy performance with AI and automation vs. manual configuration

Redefining the best network network no longer means just fast and reliable. It’s time
to redefine the ‘best network’ as a network built not just
We consider energy performance to be a triad comprised around high performance, but also with due consideration
of three elements: network performance, energy to sustainability and energy efficiency. The industry needs
consumption, and user experience. to recognize that while network KPIs are important and
As service providers focus on securing network energy must be kept in mind, the foremost priority should be
performance and achieving sustainability targets, the best optimizing networks based on delivering the wanted user
experience.

4. Figure adapted from the one used by Emmanuel Kolta (GSMA principal analyst)
Ericsson | Sustainable networks. The Intelligent Operations guide 11

Challenges for energy performance optimization: • Fixed and static settings increase operation complexity.
The data traffic in a specific radio base station site varies
• Securing performance levels while executing energy largely with daily, weekly, and seasonally variations.
saving features. Many CSPs today successfully use energy Also, the location of a site has large impact on the traffic
saving software to make significant energy and opex levels and variations. The dynamics in traffic make it
savings in their networks. Though, to preserve network challenging to use fixed settings and thresholds in energy
KPIs, in most networks the functionality is not used to its saving features to the secure both the performance levels
fullest potential. In addition, the user experience depends and to achieve optimal energy savings for each site. By
on a combination of network performance and the applying insights from network data, sites and service
applications running in the user equipment. To increase the areas can be segmented into different categories, each
energy savings with secured performance levels in the RAN with its own settings and thresholds. This will help improve
performance levels and savings. The challenge to realize
require enhanced capabilities such as observability and
these improvements lies in the enhanced operational
service awareness, with user experience in focus.
complexity as the settings require manual handling and
• Energy counters not available in old equipment. follow-up which may become an obstacle for an effective
Observability is a vital factor to optimize energy implementation.
performance and requires information about performance
• Interoperability and orchestration between different
indicators and energy use from multiple network parts.
energy-saving solutions ensuring the best selection for
The availability and accuracy of both counters and data every scenario. Ericsson and other vendors are providing a
collection pose a challenge where older equipment may great number of energy-saving solutions to support service
not have compatible counters, making pertinent data provider’s ambition to save energy. The increasing number
collection difficult with a need to complement the data with of solutions provide a large variety in how to combine
indirect measurements for instance based on estimates functionality that switch off parts of equipment, complete
of energy consumption values. Additionally, there is an frequency bands and equipment, as well different levels
inherent challenge in balancing the granularity of energy of sleep with varying reactivation times. The challenge
data. As an example, energy meters may only provide site- is further increased by the complexity of alternatives for
level understanding of energy consumption, and this level different radio access generations and frequency layers.
of granularity does not provide actionable information that Without interoperability and orchestration between the
different energy saving solutions it is difficult to ensure
can be used to effectively manage energy consumption.
the best selection of features to achieve the largest energy
savings for each scenario.
12 Ericsson | Sustainable networks. The Intelligent Operations guide

The role of benchmarking to improve energy

performance
Through performance counters it is possible to extract, for
example, data volume per site, energy consumption and
user experience, and then structure the data into different
segments. By assessing the data it’s then possible to extract
actionable insights, such as identifying areas where more
capacity will be needed and areas where more aggressive
energy-saving feature settings can be applied to achieve
larger energy savings, while still delivering a great user
experience.

Benchmarking provides context for establishing baselines,

setting realistic targets, evaluating performance and
progress. Ericsson has several data-driven collaboration
programs with service providers, such as the Ericsson
Network Benchmark. These programs use network insights
to improve network performance and energy use and provide
observability of the impact of the different actions in the
network.
• Target setting: Setting improvement targets based on
These are the building blocks for effective network energy network evolution, energy savings, and sustainability
performance benchmarking: objectives facilitates progress toward energy performance
goals, aligning with broader network ambitions.
• Observability: Identifying relevant metrics is crucial for
observing energy consumption and service efficiency, • Choice: Optimizing for energy performance while
laying the groundwork for network evolution and goal maintaining user experience is paramount, necessitating
setting. Establishing baselines allows for continuous a balance between energy efficiency and service quality.
evaluation and target setting, ensuring progress tracking Granular counters are indispensable for measuring
over time. energy consumption accurately, providing insights into
areas for upgrade and informing decisions on network
• Gaining insights: Capturing and reporting energy data
modernization strategies, including hardware and
is essential for effective energy management, enabling
software upgrades, and the deployment of AI and ML for
informed decision-making and performance evaluation.
optimization.

Identify

Observability
Targets
Choice

AI/ML-based
Measurement and
optimization
Insights transparency
with systemic trade-off

Figure 6: Energy performance benchmarking process.

Ericsson | Sustainable networks. The Intelligent Operations guide 13

04 Ericsson´s versatile
energy solutions
adapted to CSPs needs
14 Ericsson | Sustainable networks. The Intelligent Operations guide

The telecommunications industry is increasingly aware Ericsson’s holistic solution for automation relies on both
of the need to reduce energy consumption for climate and distributed real-time and centralized non-real-time
operational cost reduction. However, CSPs are hesitant automation. Moving beyond self-optimization networks
to activate energy-saving features due to concerns about (SON) architecture towards intent-driven architecture of
impacting traditional performance KPIs. Numerous software programmable networks.
functionalities allow turning off network capabilities,
including antenna branches, radios, and compute units. Service based solutions will provide a higher level of
Baseband capabilities are being enhanced to improve energy customization and shorter time to market, while platform
efficiency with energy-aware traffic steering and scheduling. evolution and software-based solutions are better
These functionalities are effective when managed and suited for a long-term evolution with improved network
controlled to minimize impact on KPIs. capabilities and simplified lifecycle management. Services
offering is evolving to rApps so that CSPs can decide
Ericsson is addressing the energy challenge for CSPs by between traditional SON architecture and the new Service
providing a range of automated energy solutions tailored to Management and Orchestration (SMO) architecture realized
their specific needs, transforming RAN energy management with the Ericsson Intelligent Automation Platform [5] (EIAP).
to be more intelligent and transparent. These solutions are There are solutions for today´s and tomorrow´s network,
designed to be future-proof, ensuring seamless evolution is the CSP choice to decide when and how to tackle on the
and adaptation of the network platform over time. This energy performance optimization.
approach allows CSPs to reduce energy consumption while
maintaining high network performance and user experience,
enabling energy performance optimization.

Predictive
Service Intelligent services
CSON Cell Energy
Continuity
Management

Energy Performance RAN Energy Site Energy

Orchestrator rApp Cockpit rApp Orchestrator rApp
SDK

Ericsson Intelligent Automation Platform

ENM 3rd party Network Resource

Non-RT RIC
EMS Management Orchestration

3rd party
EMS

5GA SW: Automated Energy Saver

RAN SW 3rd party
RAN 3rd party
Purpose-built RAN Cloud RAN RAN

Centralized Network • Network design & optimization services • SW Automation applications (rAPPs)
Automation • Network Operations & Maintenance systems • RAN domain orchestration
(Non-Real Time) • Human centric Network wide coordination • Intent driven decisions service and network actions

Distributed Radio • Rule-based Radio Resource Management

Function Automation • Al augmented Radio Resource Management
(RRM) functions
• Autonomous functions and Time critical decisions that are human monitored
(Real-Time) • High volume of decisions requiring human configured

Figure 7: Ericsson´s solutions to improve energy performance in SON and Open RAN architectures

In the figure above, we show at the left solutions in the SON type of solutions while showing our customers´
architecture and at the right in the Open RAN architecture. successful implementation.
In the following chapters, we will deep dive in both

5. Intelligent Automation Platform improves performance - Ericsson

Ericsson | Sustainable networks. The Intelligent Operations guide 15

4.1 Optus journey to improve energy

performance with automation
Optus and Ericsson have a long history of collaboration
introducing state-of-art Hardware and Software features
to realize energy efficiencies across the mobile network
to support achieving OPTUS sustainability and energy
consumption goals.

Optus and Ericsson took few steps further to incorporate

intelligent operation in relation to Energy Savings by
introducing 5G “Deep Sleep Mode Automation” into the
network.

Optus and Ericsson continue to advance their collaboration

journey by developing innovative energy saving features
in Radio Access Network (RAN) technology, starting from
“Deep Sleep Mode Automation”, continuing with Booster
Carrier Sleep Mode, Massive MIMO Sleep Mode and with
intent-based features like Automated Energy Saver in
future. This journey towards full automation and dynamic
management of energy saving framework aims to deliver
substantial RAN energy savings while improving user
experience, reinforcing OPTUS and Ericsson commitment to
sustainable and efficient network operations.

The energy challenge Solution

OPTUS has a growing focus on energy saving to achieve its Deep Sleep Mode Automation solution is a customized
goal of reducing Scope 1 and 2 emissions by 25% by 2025 solution developed to overcome this challenge. Booster
(from a 2015 baseline). Considering that 70% of network Carrier Sleep is the global solution automating deep sleep
energy consumption is from the RAN domain , there was a that Optus will be using moving forward in the ongoing
crucial need for RAN-focused energy efficiency solutions. journey towards the fully automated and dynamic energy
At the same time, OPTUS brand is focused on speed management framework.
leadership and customer experience.
This customized solution is using Ericsson’s RAN features
Balancing between these two targets cannot be simply done “5G Radio Deep Sleep Mode” and “5G Cell Locking”. It
with manual operation of energy savings features and this is aims to reduce energy consumption in 5G Active Antenna
where the challenge lies. radios Units (AAUs) as well as Remote Radio Units
(RRUs) by triggering both features during low traffic and
Although the two basic features “5G Deep Sleep Mode” and deactivating them during high traffic hours. The solution was
“5G Cell Locking” were both available since quite some time implemented across OPTUS network on one of the dominant
ago, there was a need to intelligently manage these two 5G spectrum bands (N40, or 2300MHz) in metro areas.
features and unlock their full potential for energy savings
with minimum compromise to user experience.
16 Ericsson | Sustainable networks. The Intelligent Operations guide

Results

Direct measurements of energy counters from Ericsson RAN showed a 24%

reduction in energy consumption of 5G N40 (2300MHz) layer after launching
“Deep Sleep Mode Automation” across the network.

N40 Power consumption (Wh)

17000

16000

15000
Thousands

14000

13000

12000

11000

10000

14 Jun 16 Jun 18 June 20 Jun 22 Jun 24 Jun 26 Jun 28 Jun 30 Jun 2 Jul 4 Jul 6 Jul 8 Jul 10 Jul

Figure 8: Optus energy saving results by operating intelligently

Energy consumption was reduced by 3,800 KWh/Day across CO2 emission was reduced by 2,520 KgCO2e/Day across the
the network on 2300MHz layer after introducing “Deep Sleep network on 2300MHz after introducing “Deep Sleep Mode
Mode Automation”. Automation”.

“Optus is working with Ericsson on the cutting edge technology and leveraging on automation
to continue our commitment toward energy efficiency improvement and sustainability. We are
very proud of this initiative and the contribution it will have on reducing CO2 emissions.”

Kent Wu
VP, Access Network Strategy, Planning and Quality, Optus
Ericsson | Sustainable networks. The Intelligent Operations guide 17

05 How Programmability
can make networks more
sustainable

Programmable networks offer a dynamic and flexible these networks can intelligently manage resources, ensuring
framework that allows for real-time adjustments and that energy consumption is minimized without compromising
optimizations based on current network conditions and performance or user experience. The operations team will be
demands. By leveraging advanced software and automation, empowered to achieve their targets.
18 Ericsson | Sustainable networks. The Intelligent Operations guide

Programmable networks are intent-driven

The role of an intent is to communicate requirements, Intent driven automation enables an additional abstraction
goals, constraints, and preferences to an autonomous layer in mobile networks to feed the business targets into the
system [TMForum, intent based automation]. Today system and translate those business targets into actionable
configuration parameters tell the system what to do, but configurations of the system. For example, massive MIMO sleep
by moving towards intents the operations of RAN can mode reacts to the load in the cell but the business target is
be much more efficient. Energy efficiency features today probably more concerned with the user experience which maps
primarily focus on load or other metrics that while easy uniquely to the cell-load depending on the cell characteristics.
to measure are not so interesting from a business point We call these business targets intents.
of view and break the control loop between business
The three new technology capabilities for RAN
targets and system actions. Intents allow directly feeding
programmability are service- aware, AI native, and intent
business requirements into the system which creates
driven. Without these new capabilities, the operations team
a closed loop from business targets to system actions,
can´t possibly manage multiple services guaranteeing SLAs in
simplifying both operations and analysis.
real time and consider the energy performance at the same time.

CSP defines service types and 5G RAN intents

Weeks Service Management and Orchestration

Centralized
rApps

automation
Days
via rApps
non real-time
Seconds
Intents and
Intents Observability observability

Milliseconds
Distributed
5G RAN

Intent handling User groups Observability

automation Service-aware
via RAN software RAN software
real-time Microseconds 5G RAN

Figure 9: Intent driven networks architecture in the RAN

Ericsson | Sustainable networks. The Intelligent Operations guide 19

5.1 5G Advanced software is a starting point

to programmable networks
In the RAN domain, we are making the RAN capable of In this new automation architecture, the SMO is an important
understanding CSPs intents. This shift enables CSPs to adapt element of programmability and the combination of both
the behavior of networks to achieve their business targets non- real time (rApps) and real time (radio) functionality will
while reducing operations complexity. We are building a cover all the uses cases for advanced intelligent automation
horizontal SW layer common for both cloud and purpose- with limited complexity in an efficient and sustainable way.
built RAN deployments to enable any architecture of the CSP
choice and openness. This new SW layer is named Ericsson
5G Advanced (5GA) software.

Service Management & Orchestration

Intelligent
Open management O1/A1 + R1 automation

5G Advanced software

Cloudification 5G software
with hardware Open
and software interfaces
Cloud RAN Purpose-built
disaggregation
Cloud platform
RAN Compute
General purpose HW

Open RAN fronthaul

Radio

Figure 10: High Performing Programmable Networks open architecture

Focusing on the basebands, AI-powered functionality is 5GA software is providing these new RAN programmable
capable to solve multidimensional problems with potential capabilities for new revenue generation and OPEX reduction
trade-offs, and the observability of the service performance is from energy (reducing energy consumption) and operations
a key enabler for new revenue generation with differentiated cost (with intelligent automation). This software will further
connectivity. As example: Our scheduler can make 14 billion optimize energy performance with enhancements in the
of calculations per second with the feature Interference scheduler for symbol and slot allocation optimization.
Sensing that can improve the throughput by 22%.
20 Ericsson | Sustainable networks. The Intelligent Operations guide

5.2 The three layers to improve

energy performance
There are three layers, in this new architecture, that play a node. Energy performance functionality in this layer is
role to improve RAN energy performance: implemented by Automated Energy Saver. [6]
• Energy saving RAN software features in real time • Centralized automation resides in the rApps hosted in the
enable the smart use of the network assets. This includes Service Management and Orchestration platform (SMO).
antenna branch shut off, full radio sleep, minimized use of This layer takes decisions based on neighboring node
power amplifiers between transmission slots etc. These performance to balance the overall system. It can receive
components can be viewed as complex on/off switches multiple intents and communicates with the distributed
to the lightbulb, with the ability to react to conditions like automation layer. By doing that, it can both help to predict
traffic load. Some examples are Massive MIMO sleep and local surges due to traffic migration and safeguard against
Booster Carrier Sleep. impacts due to decisions taken on neighboring sites.
Energy Performance Orchestrator is the Ericsson rApp that
• Distributed automation also resides in the RAN software,
implements energy performance optimization.
known as 5G Advanced software. It receives intents and
enables dynamic configuration and orchestration of the This architecture enables to achieve improved energy
different energy-saving RAN features in flexible time performance at scale, at network level.
while seeing the full impact on the traffic in the specific

Energy saving features to :

• Switch on/off carriers

and cells
• Reconfigure RF
channels
• Run basebands and
radios on low power Centralized (Non - RT RIC
modes automation for energy
saving Centralized
(Non-Re-al Time)
Automation

Distributed (Real Time)

Automation

Figure 11: The three layers to improve energy performance

6. rApp
Ericsson | Sustainable networks. The Intelligent Operations guide 21

Automated Energy Saver

Automated Energy Saver (AES) is a functionality After receiving the intent, AES figures out the ideal
implemented in the distributed automation layer in the 5G configuration for the energy saving features while seeking
Advanced SW. AES’ role is to orchestrate the individual to ensure the intent. This distributed architecture allows AES
energy saving functionality to fulfil an intent. access to very fast and fine granular data to prevent traffic
impact more accurately.
The intent is received in the form of wanted user experience
for a minimum percentage of user sessions (ie. X Mbps or We will stepwise introduce functionality as part of the intent-
more for at least Y % of sessions). based AES solutions, such as massive MIMO Sleep Mode and
Booster Carrier Sleep.

For AES and for other use cases, Ericsson calculates user
experience by filtering on sessions that are important for
the users, such as VoIP, consuming news, mails, videos
etc. excluding shorter and smaller sessions that are
typical of background traffic. This type of information is
part of the Ericsson´s Network benchmarking report.

Intent:
Minimum throughput per user or
95% of users with X Mbps

Automated Energy Saver

Massive MIMO Booster Carrier

Sleep Sleep

Observed
user experience

Figure 12: Automated energy saver overview

22 Ericsson | Sustainable networks. The Intelligent Operations guide

Energy performance
orchestrator rApp

The Energy Performance Orchestrator (EPO) rApp, running With centralized automation functionality, the next step
on an SMO-platform, enables centralized automation for CSPs beyond KPI guarding, will be to balance energy
to improve energy performance. This setup provides a performance and user experience across neighboring nodes,
comprehensive network view and aligns intents across guiding distributed automation loops towards the global
the network. By combining distributed and centralized optimum and ensuring energy features do not impact the
automation, operators benefit from both a broad network lower bounds of the intents.
perspective and detailed traffic insights.
The Automated Energy Saver receives as input simple
This centralized automation will handle multiple intents, intents. These intents are used to continuously adjust
ensuring they do not degrade below set limits due to energy the energy efficiency features´ parameters to meet these
efficiency features, this limits the potential downsides intents. With centralized automation, provided by EPO, more
and enables operators to pursue energy efficiency more complex intents can be managed.
aggressively.

cell cluster

KPI 1

KPI 2

KPI 3

KPI 4

Figure 13: Centralized automation in combination with distributed automation

Ericsson | Sustainable networks. The Intelligent Operations guide 23

5.3 Telstra’s commitment: Enhancing user

experience with sustainability
Telstra, in collaboration with Ericsson, is the first operator in Automated Energy Saver implementation and
the Southern Hemisphere to successfully launch Automated trial results
Energy Saver (AES) Ericsson and Telstra launch Energy
Saver for 5G network, a first-of-its-kind energy management Automated Energy Saver (AES) functionality receives intent
feature, in the commercial network. This innovative solution as an input. The intent set in this case is a minimum user
throughput for a certain percentage of users in a node. It then
enables Telstra to manage network operations based on
orchestrates and optimizes the NR Massive MIMO (M-MIMO)
defined business intents in an advanced, intelligent, and
Sleep function by dynamically finding ideal thresholds
automatic fashion. Seamless customer experiences are
for activating or deactivating antenna branches without
maintained by optimizing resources and energy consumption compromising the service quality or network performance.
on a real time basis from observed traffic flows. In the trial cluster comprising of 6 radio nodes, intent was
configured on the node to maximize energy savings from
Enabling high-performing programmable networks are
NR M-MIMO Sleep feature while ensuring at least 70 Mbps
becoming increasingly essential to unlock the full potential
downlink throughput for 90% of users. AES utilized cell and
of 5G with new advanced services, more efficient operations, node-level user experience throughput metrics to make
and enhanced customer experience. Intent-based decisions about adjusting Physical Resource Block (PRB)
automation is a key technological pillar of programmable utilization thresholds. Performance was monitored using
networks. It brings a new operations paradigm where counters for sleep time and energy consumption.
operators can use service-related metrics such as user
throughput inputs to the radio access network rather than AES was observed to maintain the set intent whilst achieving
~30% additional sleep time in the trial sites. This resulted
setting radio parameters manually. This makes AES, part of
in aggregated additional savings compared to the baseline
Ericsson’s 5G Advanced capabilities, a pivotal feature in the
NR M-MIMO Sleep feature. The highest energy saving
move toward programmable networks.

AES vs NR M-MIMO Sleep - Energy Saved (%)

0.60
4.38%

5.09% 0.3%
3.56% 2.23% 3.42%
5.5%
0.50 Highest % of 0.77% 0.62%
1.29% 0.94%
energy 0.03%
saved
4.89% 4.06%
0.40
6.83%
-0.32%
5.73% -0.07%
Energy Consumed (kWh)

0.30

0.20

0.10

0.00
Site1-1 Site1-2 Site1-3 Site2-1 Site2-2 Site2-3 Site3-1 Site3-2 Site3-3 Site4-1 Site4-2 Site4-3 Site5-1 Site5-2 Site5-3 Site6-1 Site6-2 Site6-3

Trial Cells
-0.10
AES M-MIMO

Figure 14: Automated Energy Saver vs M-MIMO Sleep – Energy saved (%)
24 Ericsson | Sustainable networks. The Intelligent Operations guide

recorded in a trial cell was approximately 6.8%. As illustrated Equally important, cells were observed to be less likely to
in the plots, these additional savings were achieved due enter sleep mode when user experience fell below the set
to AES dynamically adjusting sleep conditions whenever intent. The key finding is that periods of sleep are dictated
the average user throughput exceeded the defined intent, by the value of the configured intent and the unique traffic
thereby enabling extra sleep time. patterns/load of each node. Thus, AES simplified RAN energy
optimization by natively handling the trade-off between user
experience and energy consumption.

AES vs NR M-MIMO Sleep

100 1000

90 900

80 800

70 700

M-MIMO Sleep Time (sec)

% of users meeting intent

60 600
Cell is less likely
Extra sleep opportunity
while user experience is to sleep while
50 above set intent with user experience 500
AES+NR M-MIMO Sleep is below set
resulting in higher energy intent
40 saving 400

30 300

20 200

10 100

0
12 AM 4 AM 8 AM 12 PM 4 PM 8 PM 12 AM 4 AM 8 AM 12 PM 4 PM 8 PM 12 AM 4 AM 8 AM 12 PM 4 PM 8 PM
Day 1 Day 2 Day 3
Time

AES - Sleep Time NR M-MIMO - Sleep Time AES - Users_Above_70Mbps_% NR M-MIMO - Users_Above_70Mbps_%

Figure 15: Automated Energy Saver vs NR M-MIMO Sleep

Future collaboration between Ericsson and continues to lead in network and service innovation while
Telstra for more intelligent and energy efficient balancing energy savings with customer experience.
networks
Ericsson's strategic partnership with Telstra is vital in the
As the first 5G Advanced energy-saving feature deployed in transition towards programmable, intelligent, performance-
Telstra’s network, AES represents the start towards a more based networks. This collaboration leverages Ericsson's
advanced, intelligent, and automated network that can advanced technology and expertise in network automation
enable more sustainable operations. By deploying intelligent to support Telstra's vision of an efficient and sustainable 5G
systems that autonomously manage energy, Telstra infrastructure.

“Achieving Telstra’s ambition to reduce scope 1 and 2 emissions by 70% and scope 3 emissions
by 50% by 2030 requires a combination of decommissioning legacy and implementing new
energy efficient technology in our network. Ericsson’s Automated Energy Saver feature is an
important step on our roadmap to enabling intent driven outcomes, and in this case balancing
energy savings while maintaining a desired customer experience.”

Channa Seneviratne
Executive for Technology Engagement Advancement, Telstra
Ericsson | Sustainable networks. The Intelligent Operations guide 25

5.4 DNB is enhancing Malaysia´s 5G network

with Automated Energy Saver
Digital Nasional Berhad (DNB) serves as the sole wholesale DNB´s unique challenges
network provider in Malaysia, offering 5G infrastructure
• DNB’s NR network involves One Layer of 3.5 GHz spectrum
access to the country’s six mobile network operators
to provide NR services and B28 as the anchor layer. The
(MNOs). As the exclusive 5G network host for these MNOs, implementation of Energy saving features comes with a
it is crucial for DNB to maintain a sustainable and energy- trade off on user experience, coverage. So, the challenge
efficient network without compromising user experience. for the DNB network is unique – How can we save energy
This initiative aims to reduce energy costs and align DNB’s and reduce CO2 Emissions with less impact on Network
sustainability targets with governmental expectations. KPIs as 3.5 GHz is the only NR Layer in DNB?

DNB faced significant challenges in its efforts to lower power • Given DNB is a Multi-operator Core Network (MOCN),
consumption within its network. Primarily, reducing energy the traffic trends of each MNO is unique. Thus, finding
the right thresholds, time window/s and configuration for
usage involved trade-offs that could adversely affect user
energy saving features is almost impossible, advanced
experience and downlink throughput, which is unacceptable
automation is needed.
in a competitive 5G landscape. Additionally, determining the
optimal thresholds and timing for activating energy-saving AES implementation and results
features was complex due to daily and sectoral traffic pattern
Ericsson’s latest 5G-Advanced innovation to reduce energy
fluctuations. This variability made it difficult to minimize
costs while maintaining user experience was leveraged to
the impact on network performance while implementing
strike the balance between energy savings and optimum
necessary energy-saving measures. Moreover, continuously user experience. With the AES, it was possible for DNB to
adjusting these thresholds as traffic patterns evolved define the minimum user experience of the site. This helped
required substantial engineering resources and rigorous in simplifying RAN operations of energy optimization by
analysis to maintain a delicate balance between energy automatically recommending thresholds of energy saving
efficiency and network performance. features by natively balancing between user experience and
energy and help in maximizing energy savings.

Intent

NR 3.5 GHz M-MIMO Sleep Automated Energy Saver

Observed
User experience

Figure 16: Simplified Network settings enabled by AES

26 Ericsson | Sustainable networks. The Intelligent Operations guide

With the AES feature DNB was able to have up to 4.8% AES is able to do this because it continuously evaluates the
average Power savings in the site where AES was best parameter configuration in real time, improving network
implemented. The highest power savings goes up to 7.4% on adaptability while reducing operational efforts.
few sites. It was observed that there were opportunities to
On the sites where AES was implemented, the result shows
save energy throughout the day, and that the AES was able
that average of 1.531 kWh were saved per site during the
to adjust the thresholds automatically within the thresholds
period of activation.
set for user experience.

2.7%
4.2%
3.6%
6.5% 5.3%
3.2% 7.4% 4.3%
4.7%
6.5%

Site 1 Site 2 Site 3 Site 4 Site 5 Site 6 Site 7 Site 8 Site 9 Site 10

Before AES (kWh) Before AES (kWh) Before AES (kWh)

Figure 17: Trial results on AES enabled sites

Automated Energy Saver is the first 5G Advanced feature operate intelligently and save energy with no impact on user
deployed in DNB’s network. With AES DNB has started the experience.
transformation towards an intent driven RAN network to

“I am proud to see DNB’s 5G network continuing to be a pioneer in the region by being the first
in Southeast Asia to deploy the functionality of the Automated Energy Saver (AES), which
is just one of the many 5G Advanced capabilities supported by our network. The advanced
automation will allow us to monitor and manage user experience while granting us the ability
to boost energy savings with reduced operational complexity. This means the network will be
sustainable by being energy efficient while maintaining undiminished network performance.”

Ken Tan
Chief Technology Officer, DNB
Ericsson | Sustainable networks. The Intelligent Operations guide 27

5.5 Swisscom 5G journey commitment

to performance and sustainability
Swisscom and Ericsson have partnered to deploy 5G network Swisscom is committed to achieving net zero emissions by
in Switzerland [7]. 5G is the most efficient technology 2035 achieving a carbon-neutral value chain and reducing
and 5G Standalone architecture (SA) maximizes energy energy consumption in its infrastructure by implementing
efficiency. In April 2019, Swisscom launched commercial energy-efficient technologies. This commitment aligns with
5G in Europe, achieving numerous milestones with Ericsson. Swisscom’s recognition of setting ambitious goals to reduce
Today, Swisscom provides 5G coverage to 96% of the Swiss emissions across its operations.
population and continues to expand. In May 2021, Swisscom
The combination of both targets resulted in the decision
and Ericsson committed to a 5G Standalone (SA) rollout,
to implement Automated Energy Saver functionality in
enhancing network performance, and introducing new use
Swisscom’s RAN to ensure both performance and zero
cases.
energy waste.

7. The Swisscom 5G story - A vision for 5G - Ericsson

28 Ericsson | Sustainable networks. The Intelligent Operations guide

Automated Energy Saver test results

49%
As part of its sustainability drive, Swisscom, in collaboration
with Ericsson [8], conducted field tests on the Automated increase time on sleep
Energy Saver (AES) feature within its RAN network. This AES mode in low traffic
feature, designed to enhance energy efficiency, optimizes
power consumption by implementing a sleep mode in
NR Massive MIMO configurations without compromising
network performance.

The test was designed to compare the energy savings of

NR Massive MIMO in it’s default configuration with NR

2%
Massive MIMO augmented by the Automated Energy Savers
algorithm for activating/deactivating the functionality. cluster energy savings in the
The key performance indicators (KPIs) used to evaluate trail securing KPIs
AES effectiveness included pmConsumedEnergy and
pmMassiveMimoSleepTimeTxMute, keep track on energy
consumption and the active duration of the sleep mode,
respectively.

The test results demonstrated a significant achievement in Future Rollout

energy efficiency. During AES activation, the sleep mode
active duration increased by around 49%, highlighting the The test results align closely with Swisscom’s sustainability
system’s enhanced responsiveness and energy conservation goals by providing measurable energy savings while
during lower traffic periods. AES feature yielded a 2% maintaining network integrity. Based on these positive
reduction in energy consumption across the cluster, with even outcomes, Swisscom plans test AES feature in a larger area of
higher savings observed on nodes with heavier traffic loads. its network, marking another step toward a greener and more
efficient telecommunications infrastructure. The anticipated
Moreover, AES became a shift to more dynamic handling wider rollout of AES, expected in 24Q4, represents
of Massive MIMO Sleep Mode thresholds further supported Swisscom’s continued investment in sustainable innovation
Swisscom’s sustainability goals by adapting to real-time and energy-efficient solutions within its operations.
network demands.

“We’ve been working closely with Ericsson for over 10 years with a great amount of trust and
success. This will enable us to not only offer our customers the best customer experience, but
also to place an even greater focus on sustainability and innovation.”

Gerd Niehage
CTIO Swisscom

8. Swisscom and Ericsson – strategic partnership | Swisscom

Ericsson | Sustainable networks. The Intelligent Operations guide 29

06 Services powered with

AI to maximize energy
savings

While AI can boost energy performance, its execution brings Our service solutions reduce barriers to boosting energy
new challenges for the CSPs. performance using AI on Ericsson´s platforms to execute
AI-based models and algorithms. In this chapter, we will
AI implementation complexity challenges are:
explore some of these solutions that can be applied in
• Serving RAN data to ML, efficiently and sustainably. Ericsson and multi-vendor radio networks. These service-
• Understanding DataOps and use cases. based solutions offer CSPs a unique opportunity for
customization based on specific and ad-hoc needs.
• Resource limitations and latency requirements.
• Data locality, hardware-optimized models, managing the AI based automation provides additional gains with data-
lifecycle of the model. driven self-management. The system will be able to gather
network data and update the existing configuration without
human intervention, this will release the operations team
from this time consuming updating task.
30 Ericsson | Sustainable networks. The Intelligent Operations guide

6.1 Intelligent RAN Power Saving solution

with a tailored approach
Intelligent RAN Power Saving solution is part of the Service (Booster Carrier Sleep). These features require varying
Continuity suit also known as AI-apps. This service aims activation times, with hardware features needing a few
to pre-empt and predict events in CSP networks to act seconds and software features requiring less time.
before they occur. This approach is also applied to energy
The algorithm is being refined through the experience
efficiency. The energy efficiency algorithms (AI-APPs) use
gained from multiple networks. With over 10 customer cases
AI/ML to predict network traffic and activate or deactivate
globally in trial and commercial deployments
radio hardware features (Deep Sleep) and software features

Resilence
and
simplicity

Co-creation from the get-go

Either choose from o ready-to-use Al-apps or get
to create your own. We merge network-specific
Barrier-free start data from you and already aggregated dato sets
We remove the entry barriers for your Al ot hand, allowing to odapt for chonging inputs
journey. Your service bundle is based on over time.
value-and TCO- colculations and allows you
to start AL integration without huge
monetary commitment.

Scope of engogement

Figure 18 : AI journey made easy with network services

How it works can switch off all digital signal processing parts of the radio,
achieving up to 90% reduction in energy consumption
Coverage and target cell concepts in a pre-deployment data compared to average radio nodes.
study based on PM counters are extracted from CSP network
and analyzed by dedicated Data Scientist Team. Output of When traffic is predicted to return to normal levels, the
this is the classification of cells and the input to ML algorithm algorithm will wake the node and resume normal activity.
training. In this way the resulting artifact is tailored on The algorithm also monitors the impact on cells that remain
CSP network by definition: in fact, data are inherently active, ensuring they can handle the traffic shifted from the
warrantying the customized service approach. sleeping cells. It is crucial to monitor the energy consumption
of these active cells to ensure overall energy savings.
Once the algorithm is ready, it will be deployed to
continuously compare real-time network data with traffic All these functions are managed by the Ericsson services
predictions from the ML algorithm. If the algorithm predicts platform, where the application/algorithm runs. The platform
energy savings and the network data confirms low traffic, it is essential for this service delivery, connecting directly to
will send a command to put the node circuitry to sleep. This nodes to gather performance counters and configuration
technology, embedded in Ericsson's silicon-based processors, data, and also receiving configuration information from the
Network Management layer, such as the Element Manager.
Ericsson | Sustainable networks. The Intelligent Operations guide 31

6.2 Umniah AI-based solution to improve

energy efficiency in Jordan
Since its launch in 2005, Umniah, a subsidiary of Beyon reduction in energy consumption in network operations
Group, has been recognized as one the fastest growing through its smart Power Saving AI apps while securing
and reliable telecommunications providers in the region’s network performance.
most competitive markets, offering a wide variety of
This partnership to deploy Ericsson’s cutting-edge AI/
transformative high-quality mobile, Internet, and
ML solutions will significantly reduce energy consumption
enterprise solutions.
across Umniah’s network operations in Jordan and deliver
Umniah has partnered with Ericsson to deploy Service a remarkable advancement in the telecom industry’s efforts
Continuity functionality, which is based on Artificial towards environmental sustainability.
Intelligence and Machine Learning capabilities enabling
32 Ericsson | Sustainable networks. The Intelligent Operations guide

Intelligent RAN Power saving solution it results in reductions in carbon dioxide emissions and
implementation and results simplifies operations.

The solution uses a machine learning prediction model that The solution was deployed in the live commercial network
continuously analyzes real-time network data. Through reducing energy consumption without impacting user
intelligent closed-loop decision-making capabilities, it experience.
determines whether to deactivate, activate, or maintain
The deployment comes after a successful proof-of-concept
network component-controlling parameters based on the
(PoC) where Ericsson’s Intelligent RAN Power Saving
data and activity in neighboring cells. This enables precise
solution, part of Ericsson’s Service Continuity AI app suite,
energy management and operational efficiency without
demonstrated about 20 percent on 5G daily power saving
compromising network or service performance. Additionally,
capabilities.

Based on current Data preparation

state of the feature, with historical and
the action is taken to real-time data
activate or deactivate

The model decides

if the feature shall Create behavioural
be active or not model pre cell in
the network

In real time when

data is generated,
run the data
through the model

Figure 19: Closed loop implementation for Intelligent RAN Power Savings

“The collaboration with Ericsson signifies a crucial step forward in our commitment to
sustainability and technological innovation. The positive outcomes of the initial proof-of-
concept were clear, and we are eager to see the benefits of this AI-enhanced power-saving
technology on a larger scale. The implementation of Ericsson’s Service Continuity Power Saving
solution is not just about cost savings—it is about taking meaningful action towards reducing
our environmental footprint and building a greener future.”

Alaa Ibrahim
Chief Technical Officer, Umniah Jordan
Ericsson | Sustainable networks. The Intelligent Operations guide 33

6.3 Energy Efficiency Map

This solution provides an energy efficiency map to efficiency is managed. Energy efficiency is a way not only of
understand the energy efficiency both at network level reducing OPEX but also to improve performance and drive the
and radio node level granularity. Applying AI clustering faults fix process in a preemptive way, before an alarm is raised
technology, the sites are classified in different energy by the nodes.
efficiency ranks. It monitors the energy efficiency and
identifies autonomously the root cause of network
inefficiencies.

This solution has two deployment options: as a tailored

algorithm (Node Radio Power efficiency map) in the service
continuity offering or as an rApp (Ericsson RAN Energy
Cockpit) deployed in the Ericsson Intelligent Automation
Platform (Ericsson implementation of the O-RAN Service
Management and Orchestration platform).

This application follows several steps to improve energy

efficiency:
• Visualize the overall network energy efficiency with
Cell energy efficiency status visualized in a defined rank
site granularity – not only is site energy consumption
considered, but also GB/kWh and other traffic metrics
(such as data volume and average users/cell).
• Understand the root cause of inefficiencies – this
is an automated process to identify and isolate the
cause. configuration issues impacting performance and
209
Cells
consumption, and hardware and software issues.
• Provide recommendations for resolution – once the
cause has been identified, it is then shown in the web
portal to alert the monitoring team, who will decide Rank 6 Rank 5 Rank 4
the next steps for the resolution such as implementing Rank 3 Rank 2 Rank 1
the recommendation or sending the actionable
recommendation to the networks’ operations team or
opening a Trouble Ticket. For some scenarios, automatic Figure 20: Energy efficiency map at site level. Cell energy
resolution (closed loop) can be enabled such as parameter efficiency status visualized in a defined rank
configuration.

The application is able to produce as output not only a map

The main benefits are:
of efficiency distribution through the network, but also a
report in which results of fault isolation scripts are reported • Reduced OPEX with less operations cost and reduced
and overall recommendations toward the CSP Operation time to resolution.
team are delivered through a web page dedicated to CSP • Optimized CAPEX since it identifies over-dimensioning
Operational team where OPIs and other indications are scenarios
reported.
• Improved network performance with the actions to
Node Radio Power Efficiency Map service or Ericsson RAN fix network issues identified there are both energy
efficiency and performance improvements.
Energy Cockpit rApp changes the way in which energy
34 Ericsson | Sustainable networks. The Intelligent Operations guide

6.4 Predictive Cell Energy Management

Solution for multivendor RAN
Predictive Cell Energy Management (PCEM) is an innovative advanced AI techniques, PCEM dynamically establishes
solution that minimizes radio access network (RAN) energy optimal thresholds that conserve energy without impacting
consumption across multivendor environments. Using network performance.

PCEM AI Cases and Output

Band C Window N Hours

C181 Night 8-7

Analytical Store W1 WL W3
C181 Day 8-20
(Time Window)
Data Input C181 Evening 21-23 0 C1 C2 23
W1 W3 W2
N B Cell exclusion
Network Geography
(NW, Region, Residential)

RRC High
Profile Configuration Sector name Traffic
Low Profile Configuration
Cell & Site DB Clustering Store H C1-B1-W1 Xyz S1
H C1-B1-W1 - H
(Cell name, Site name, sector name, band
name & config) (Traffic Profiling) M C1-B1-W1 Xyz_51 Action Scheduler
M C1-B1-W1- M
Med PRB
L C1-B1-W1 Xyz 51
L C1-B1-W1 - L
Throughput

CF 2.0 Implementation
Throughput= f(RRC; PRB)'

Configuration Threshold CAP

PM Data C1-B1-W1 - H T1
(RRU name, RRU model, cell Threshold Optimizer COV
configurations) C1-B1-W1- M T2

C1-B1-W1 - L T2
T1

CM Data

Cell name Energy Saving

Prediction Engine operation

Figure 21: PCEM Solution overview

The more consumption of traffic is there on the network, the The higher the network traffic consumption, the greater
more will be the energy requirements. The measurement the energy requirements. PCEM measures its effectiveness
of the treatment by PCEM is done by comparing the energy by comparing the energy consumed by the network to the
consumed by the network with respect to the data volume it data volume provided to customers. PCEM reduces energy
is providing to the customers. In simple terms, PCEM lowers consumption by deactivating unnecessary equipment during
the consumption by switching off the unnecessary equipment specific time windows without impacting data traffic. The
at certain time windows without affecting the data traffic next figures illustrate that for each data point the energy
consumption in the network. The below shown figure consumption (in kWh) versus data traffic (in Gb), both the
for a non-Ericsson vendor (Fig 8a) and Ericsson (Fig 8b) absolute consumption and the slope are lower when PCEM
shows that for each data point on a graph showing energy is active in the network compared to the baseline behaviour
consumption (in kWh v/s data traffic (in Gb), the absolute observed in the week before deployment.
consumption as well as the slope is lower when PCEM is
working in the network compared to the baseline behaviour
which is the last week before the deployment or trial.
Ericsson | Sustainable networks. The Intelligent Operations guide 35

Energy vs traffic volume

54

53
y=0.0356x+40.398
52

51

50
y=0.0356x+37.608
49

48

47
275 295 315 335 355 375 395

Energy vs Avg. PRB_Util_DL

54

53
y=0.8522x+40.947
52

51 without PCEM
50
y=0.6745x+38.564
49
with PCEM

48

47
12.0 13.0 14.0 15.0 16.0 17.0 18.0 19.0

Figure 22: Graphs showing energy consumed versus traffic volume. Reduced energy
consumption after PCEM deployment with non-Ericsson and Ericsson radios

PCEM has been deployed in multiple countries/regions for various

customers and we have gained satisfactory results. The below table
shows some of the results in the field

Region of Energy saving

Technology Vendors
deployment (per day over network)

Brazil 4G, 3G Ericsson 4.6%

UK 4G Other vendor 5.9%

Croatia 4G, 5G Ericsson 5%

Figure 23: Energy savings with PCEM in multivendor RAN

36 Ericsson | Sustainable networks. The Intelligent Operations guide

6.5 TDC NET deploys Predictive Cell Energy

Management for energy performance
TDC NET is a leading mobile infrastructure provider in The solution
Denmark with a strong commitment on the climate change
and have partnered with Ericsson with an ambition on Reducing the energy consumption of the RAN is a complex
technology leadership and enabling the green transition optimization task due to the dynamic nature of site
build, traffic patterns and the underlying importance on
towards sustainable future.
maintaining the required KPI performance of the network.
The challenge: Energy savings without impacting The PCEM solution enables TDC NET to set the desired
network KPIs network priorities, based on which the AI modules can
identify the best energy saving action to take on a specific
As part of the plans towards Science Based Target initiative cell at a specific moment in time. The solution considers
(SBTi) validated 2030 net zero target, TDC NET is taking the historical trend and traffic analysis to maximize the
actions to reduce the energy consumption in the RAN, while energy savings, which is then complemented with real-time
maintaining superior customer experience. Beside expanding
monitoring of the network KPIs to avoid any impacts to the
the mobile network with the most recent energy efficient
customer experience.
hardware, TDC NET has decided to implement Ericsson’s
Predictive Cell Energy Management (PCEM) solution which,
through AI and automation, is set to achieve higher energy
efficiency in the RAN while maintaining the expected KPI
performance of the network.
Ericsson | Sustainable networks. The Intelligent Operations guide 37

The result leveraging on AI technology and its prediction capabilities.

It is estimated that with the current scope of PCEM
The deployment of PCEM was implemented initially at 4G implementation, TDC NET would be saving ~800MWh/year
frequency layers across 58% of the network. We observed a across the targeted sites, expect to lead to a reduction of
reduction in energy consumption when transmitting similar ~135Tons of CO2e for 2024.
data volumes in the network, estimated as energy efficiency
improvement of 6% kWh/GB for the addressable scope. Further action is needed towards higher reductions of
Crucially, these reductions were achieved with no visible the energy consumption and achieving the sustainability
impact on network performance and customer experience. goals. TDC NET will continue the efforts in enhancing the
initial results and exploring the AI approach on energy
PCEM solution is enabling a reduction of energy efficiency features in the RAN to further improve the energy
consumption without impacting the network performance by performance and thus the CO2e footprint of the network.

4G PCEM Sites-Downlink Data Volume v/s Energy Consumption

127000

126000
Region of similar Data volume

125000

124000
Energy consumption [kWh]

123000

122000 6% efficiency
in kWh/GB

121000

120000
1250 1350 1450 1550 1650 1750 1850
Downlink Data volume [TBs]

Baseline Linear Baseline

Go-Live Linear Go-Live

Figure 24: PCEM results in TDC NET

“We are very pleased to continue our strong mobile partnership with Ericsson to build out our
leading position in the market and ensure the country’s best network experience. By testing
and deploying AI and automation-based solutions, we aim to enhance energy efficiency and
user experience in our network operations”.

Peter Søndergaard Andersen

Vice President, Head of Sustainability, TDC NET
38 Ericsson | Sustainable networks. The Intelligent Operations guide

07 Site solutions
innovations beyond
energy efficiency

RAN Site portfolio in Ericsson is dedicated to enabling In the transition of the power grids, enlarging the energy
sustainable infrastructure for cell-sites especially focusing reserve capacity is imperative to ensure stability, making
on energy, energy storage, interconnect solutions, security the balancing of every electricity grid the new norm. For
and renewables. Each of these areas have dedicated product power grid owners to meet this they can receive support with
line-ups within Ericsson to also enable Open RAN infra. various energy ancillary services from stakeholders in the
energy ecosystem.
Utilities companies are under big pressure to be able to
answer the energy demand. For this reason, they are
targeting mobile network operators to support in their
How can energy provider manage the grid to be
challenges, they are offering benefits as reward, these
in balance?
benefits are either special prices for energy or rewards. Energy provider can today control the balance by two
Energy consumption comes as an opportunity to join these variants, price or/and incentives. With price the energy
incentive programs. provider can steer the consumption to be very expensive
during high demand hours and with incentives (ancillary
Our solutions go beyond energy efficiency since they enable
services) motivate energy consumers to avoid consuming at
CSPs not only to reduce the energy cost but also to gain new
high demand hours.
revenue, becoming an energy player and participating in the
energy market. Telecom network site owners, who own and control the
passive equipment at the site with energy storage capability
Grid challenges including renewables, can utilize the energy storage
possibilities to avoid consumption from the grid when the
An increase in the use of electric energy source is stepping
grid has high demand, and they can also be part of the
up demand on the power grid. The ongoing transition to
energy ecosystem and participate in the various energy
renewable energy sources, which vary in output, makes
ancillary services.
it more important to match power generation with
consumption. Networks need to minimize its energy consumption while
maximizing the use of stored energy at network sites
With the shift to fossil-free energy generation and
including renewables, which requires a transition toward
electrification, power grid owners also need support to meet
an intelligent energy setup and an integrated approach to
new challenges emerging from this ongoing transformation
energy management and orchestration.
in energy generation and consumption.

Main forces challenging grid robustness are: How can CSPs reduce the overall energy
• Bottlenecks in electrical transmission capacity consumption with Smart Site solutions
• Growing share of intermittent energy sources such as solar Ericsson OPEX features in commercial operation show
photovoltaic/wind globally significant reductions in energy bills through deploying
• Reduction in the share of rotating mass for electricity methods like peak shaving, load-shifting features and using
balancing in the power grid. hybrid energy with onsite renewables and lithium batteries.
Ericsson | Sustainable networks. The Intelligent Operations guide 39

Peak shaving solution consumption exceeds threshold defined by energy provider.

With peak shaving CSPs can avoid higher tariff and peak by
Depending on the market, some countries have huge taking the power from the energy storage during high load
difference between the high price and low price on the hours and by that reduce OPEX.
tariff for electricity. Some has also higher tariff if peak

Power

Overview
• Peak shaving limits power outtake from grid
• Remaining power is drawn from battery

Benefits
• Helps reduce peak power/demand charges
Time
from utilities
00:00

01:00

02:00

03:00

04:00

05:00

06:00

07:00

08:00

09:00

10:00

11:00

12:00

13:00

14:00

15:00

16:00

17:00

18:00

19:00

20:00

21:00

22:00

23:00

• AC mains breaker size can be minimized

Power from grid Power from battery Peak shave limit

Figure 25: Energy peak shaving solution

Load-shifting features from the energy storage when the electricity tariff is high and
reload the energy storage when the electricity tariff is low,
Another way to reduce OPEX is to shift when CSPs need to can in some markets give a good reduction. This is when the
consume energy from the grid. If consumption can be taken solution is running on peak load shifting.

Tariff low Tariff high Tariff low Overview

• Load shifting disconnects entire RAN load
from grid and operates on battery at defined
time(s)
Battery operation Battery charging

Mains operation Mains operation • Battery recharge can be schedule for low or
very low tariff periods

Start peak Stop peak load

load shifting shifting Benefits
Start battery
Stop battery • Avoids high traiff periods
charging
charging

Figure 26: Energy load shifting solution

40 Ericsson | Sustainable networks. The Intelligent Operations guide

What are the different strategies followed by Site Energy Orchestration AI based software can be provided
CSPs to reduce energy cost as a service part of the service continuity suit. In the short-
term, it will be provided as an rApp deployed on the EIAP
CSPs have always the need to make sure the network is up [10] (Ericsson´s implementation of the SMO).
and running to the lowest cost, therefore cost reduction is
always on the agenda. Peak shaving or Load-shifting give Site Energy Orchestration solution is built of a combination
the CSPs option to reduce energy cost in a scheduled way but of machine learning (M/L) and AI applications and acts as an
with Site Energy Orchestration CSPs will have better options interface between global networks and the energy grids. The
to not only reduce cost but also gain revenue based on the solution gives service providers cost control, to reduce energy
load consumption in their network. OPEX without affecting the user experience and incentive,
giving service providers a possibility to participate in various
Beyond energy efficiency: Site Energy energy ancillary services.
Orchestration The solution enables service providers to cluster hundreds or
Ericsson’s Site Energy Orchestration (SEO) [9] is designed thousands of sites and orchestrate the network sites in the
to support our customers in addressing the new challenges best way by reduction of energy or participating in ancillary
that come with the transformation in energy generation and services. It also enables a combination of programable
consumption. freedom depending on the difference in the utility market
regulation. CSPs with network crossing through country
boarders will be able to adjust the solution according to the
regulation in the country.

Power grid Energy exchange market Energy spot market Weather forecast

Site Energy Orchestration AI

RAN data

Energy Renewable Solution Energy Management Solution

Figure 27: Site Energy Orchestration solution

9. Intelligent Site Energy Orchestration Solutions - Ericsson

10. Ericsson Intelligent Automation Platform
Ericsson | Sustainable networks. The Intelligent Operations guide 41

Results in the field

Thanks to the deployment of this solution a CSP in north The smart Energy site trials in the north america market
America could lower their cost significantly by applying peak demonstrated more than 20% daily savings in energy costs
load shifting feature both in summer and winter time. When
adding peak shaving in combination with peak load shifting,
OPEX is further reduce.

Cost with Cost with

Load Shifting only Load Shifting+ Peak Shaving

-21% -26 %

-15% -18 %

Energy cost Demand charges

Figure 28: Smart Energy Site solution results in a North America Mobile Network
42 Ericsson | Sustainable networks. The Intelligent Operations guide

08 Key takeaways and

recommendations

Key takeaways

New software functionalities will allow CSPs to efficiently scale-up or

Enable energy scale-down network capacity to the traffic load. High-performing, energy
performance efficient radio network equipment can dynamically adapt with the finest
optimization with granularity to prevent any energy waste. Ericsson´s tailor-made silicon is the
foundation of our high-performing radio network with yearly improvements
high-performing radio
on energy efficiency, enabling better sleep modes, and increased processing
networks capabilities.

Networks operations shift towards intent-driven networks will bring

Intent-driven multiple benefits, including improved user experience and reduced energy
automation with AI will consumption. This new operations paradigm will allow CSPs to simply
secure user experience state their multiple objectives or intents (i.e securing user experience and
reduce energy consumption) while the RAN handles complex processes and
and reduce energy
actions. AI technology will help in high-complexity scenarios evaluation with
consumption multiple optimization objectives and potential trade-offs.

The introduction of differentiated connectivity with performance-based

Evolving to services is a game changer for operators, securing user experience while
programmable optimizing energy consumption. Ericsson’s vision of programmable networks
networks will help CSPs represents a holistic approach to energy performance, embedding it into every
use case and scenario. This strategy addresses the immediate need for reduced
to achieve business and
energy consumption and secured user experience as week as paves the way
sustainability targets for a more sustainable and adaptable network infrastructure in the future.
Ericsson | Sustainable networks. The Intelligent Operations guide 43

Actionable recommendations for CSPs

Define new metrics to redefine the best network

01
Ericsson believes that it is time for the industry to redefine what is considered as the best network,
moving beyond the traditional metrics of speed and coverage, and instead focusing on developing
networks that are high-performing, energy-efficient and sustainable.

Enable holistic observability across the entire network

02
Observability is a vital factor to optimize energy performance and requires information about
performance indicators and energy use of different radio technologies from multiple network parts.
This includes active telecom equipment, such as radios and baseband units, as well as from site-
support systems, such as rectifiers and cooling units. This information is typically available from
data counters through embedded systems in the network. This holistic view is important because
focusing on radio performance alone is insufficient to accurately assess the energy performance of
the mobile network.

Improve energy performance with benchmarking

03
The introduction of differentiated connectivity with performance-based services is a game changer
for operators, securing user experience while optimizing energy consumption. Ericsson’s vision
of programmable networks represents a holistic approach to energy performance, embedding it
into every use case and scenario. This strategy addresses the immediate need for reduced energy
consumption and secured user experience as week as paves the way for a more sustainable and
adaptable network infrastructure in the future.

Leverage your energy assets to reduce costs

04 and gain new revenue
Mobile network site owners, who control passive equipment with energy storage capabilities,
including renewables, can use these storage options to avoid grid consumption during high
demand periods. They can also join the energy ecosystem and participate in various ancillary
services. CSPs can turn energy challenges into opportunities by becoming energy players.

Secure interoperability and orchestration between

05
different energy-saving solutions
Ericsson and other vendors are providing a great number of energy-saving solutions to support
service provider’s ambition to save energy. The increasing number of solutions provide a large
variety in how to combine functionality that switch off parts of equipment, complete frequency
bands and equipment, as well different levels of sleep with varying reactivation time. Without
interoperability and orchestration, it is difficult to ensure the best selection of features to achieve
the largest energy savings for each scenario. Ericsson´s proposed architecture with 5G Advanced
software, SMO and rApps ensure both orchestration and interoperability between different
functionality and vendors.
44 Ericsson | Sustainable networks. The Intelligent Operations guide

Empower your operations team with AI-based

06
automation solutions
Network complexity is further driving the need to use automated solutions with AI/ ML to operate
the networks. Communications Service Providers are keen to improve energy performance in the
active and passive equipment without impacting user experience and network resilience. AI based
automation provides additional gains with data-driven self-management. The AI system will be
able to gather network data and update the existing configuration without human intervention,
relieving the operations team from this time-consuming updating task. Ericsson can provide both
services and software-based solutions.

Identify what use cases can benefit the most

07
from AI technology
High-complexity scenarios with multiple optimization objectives and potential trade-offs require
the use of AI. The predictive capabilities of AI based solutions can boost energy savings without
compromising user experience. In the CSPs’ AI journey, it will be essential to focus on leveraging AI
whenever gains are more significant and legacy solutions can be seamlessly replaced to maximize
the return of investment (ROI

Evolve networks to be intent-driven and programmable

08
Programmable networks can manage the complex interaction of assets, site-specific
optimizations, and ongoing maintenance by reducing the operational effort with the intent
definition at network level. Intents abstracts the complexity of underlying processes and allows
CSPs to interact with the system more intuitively telling “what to achieve” instead of the “how”.
This shift will also enable performance-based business models in a cost-efficient way.

Intent-driven and AI powered networks can provide the most optimized solution for multi-
objective targets such as securing user experience and optimize energy consumption. Combined
centralized and distributed automation will manage multiple objectives or intents. Ericsson is
pushing the technology boundaries for a conscious energy use and user experience targets.
Ericsson’s high-performing, programmable networks provide
connectivity for billions of people every day. For nearly 150 years,
we’ve been pioneers in creating technology for communication.
We offer mobile communication and connectivity solutions for
service providers and enterprises. Together with our customers and
partners, we make the digital world of tomorrow a reality.

ericsson.com The content of this document is subject © Ericsson AB 2024

to revision without notice due to
continued progress in methodology,
design and manufacturing. Ericsson
shall have no liability for any error or
damage of any kind resulting from the
use of this document